Reflexivity in Economics 2009

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/268024421
Reﬂexivity in economics. An experimental examination on the self-

referentiality of economic theories
Book · January 2009

DOI: 10.1007/978-3-7908-2092-8
CITATIONS READS
22 375
1 author:
Serena Sandri
German Jordanian University
33 PUBLICATIONS 300 CITATIONS
SEE PROFILE
All content following this page was uploaded by Serena Sandri on 17 April 2020.
The user has requested enhancement of the downloaded file.

CONTRIBUTIONS TO ECONOMICS
Serena Sandri
Reflexivity
in Economics
An Experimental Examination on
the Self-Referentiality of Economic
Theories
Physica-Verlag
A Springer Company
Reflexivity in Economics
Serena Sandri
Reflexivity in Economics
An Experimental Examination on the
Self-Referentiality of Economic Theories
Physica-Verlag
A Springer Company
Dr. Serena Sandri
Humboldt-Universität zu Berlin
Wirtschaftswissenschaftliche Fakultät
Institut für Entrepreneurship
Spandauer Str. 1
10178 Berlin
Germany
[email protected]
ISBN: 978-3-7908-2091-1 e-ISBN: 978-3-7908-2092-8

DOI: 10.1007/978-3-7908-2092-8
Contributions to Economics ISSN: 1431-1933
Library of Congress Control Number: 2008935108
© 2009 Physica-Verlag Heidelberg
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer. Violations
are liable to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
Cover design: WMXDesign GmbH, Heidelberg
Printed on acid-free paper
9 8 7 6 5 4 3 2 1
springer.com
To Ali, my father-in-law
Acknowledgments
I would like to thank my supervisor, Prof. Marco Lehmann-Waffenschmidt, for his

support and helpful comments.
Further, I especially thank Prof. Werner Güth for his constructive suggestions
and for his supervision of the experimental part of this work.
For funding the experiments reported in this dissertation I gratefully acknowl-
edge the Max Planck Institute for Economics in Jena and the University of Bari.
vii
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 Reflexivity and Self-Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Reference Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1 Reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.2 Universality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.3 Ungroundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3 Varieties of Self-Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Taxonomies of Self-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Self-Reference in Natural Language . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.6 Self-Reference in Formal Language . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7 Logical Consistency of Self-Reference . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7.1 Self-Reference and Paradoxes . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.7.2 Harmless and Harmful Self-Reference . . . . . . . . . . . . . . . . . . . 24
1.8 Reflexivity in Human Understanding . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.9 Self-Reference in Social and Individual Decision-Making . . . . . . . . . 26
2 Reflexivity of Social Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.1 What is Social Reality? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Recursivity of the Social Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.1 Reflexivity in Anthropology . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.2 Reflexivity in Linguistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2.3 Reflexivity in Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2.4 Reflexivity in Politics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.5 Reflexivity in Sociology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.2.6 Reflexivity in Psychology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.2.7 Reflexivity of Economic Reality . . . . . . . . . . . . . . . . . . . . . . . . 44
ix
x Contents
3 Reflexivity and Predictability of the Social Sciences . . . . . . . . . . . . . . . . 53

3.1 Constructs and Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.1 Observer, Observation and the Construct of the Self . . . . . . . 57
3.1.2 A Constructivist Approach to the Cognitive Processes . . . . . 59
3.1.3 Scientific Research and Reflexivity . . . . . . . . . . . . . . . . . . . . . 62
3.1.4 Science as Language Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.1.5 Constructivism and Economics . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.2 Recursivity of Social Theorizing and Predictability
of Social Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2.1 Social Predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2.2 Explaining and Predicting the Social Reality . . . . . . . . . . . . . 69
3.2.3 Reflexive Predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4 On the Rationality of the Economic Actors . . . . . . . . . . . . . . . . . . . . . . . . 77

4.1 Questioning the Descriptive Validity of Rational Choice
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1.1 The Neoclassical Defence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.1.2 Allais’ Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1.3 Simon’s Bounded Rationality Approach . . . . . . . . . . . . . . . . . 81
4.2 The Bounded Rational Revolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2.1 Adaptive and Satisficing Behaviour . . . . . . . . . . . . . . . . . . . . . 84
4.2.2 Principles of Problem-Solving . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2.3 Problem-Solving in Games and Puzzles . . . . . . . . . . . . . . . . . 89
4.3 Decomposing Rationality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.3.1 Intuition and Reasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.3.2 Accessibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.3 Framing Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3.4 Prospect Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5 Heuristics, Biases and Methods for Debiasing . . . . . . . . . . . . . . . . . . . . . 101

5.1 Bounded Rational Heuristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.1.1 Building Blocks of Bounded Rational Heuristics . . . . . . . . . . 103
5.1.2 Main Features of Bounded Rational Heuristics . . . . . . . . . . . . 104
5.2 Heuristics and Biases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.2.1 Representativeness Heuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.2.2 Availability Heuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2.3 Adjustment and Anchoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.3 Debiasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.3.1 Debiasing the Representativeness Heuristics . . . . . . . . . . . . . . 112
5.3.2 Debiasing the Availability Heuristic . . . . . . . . . . . . . . . . . . . . . 114
5.3.3 Debiasing Adjustment and Anchoring . . . . . . . . . . . . . . . . . . . 114
5.4 Concluding Remarks on Debiasing and Some Implications
for Theory Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Contents xi
6 Self-Referentiality of Economic Theories and Theory Absorption . . . . 119

6.1 Economic Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.2 Self-Referentiality of Economic Theory and Theory Absorption . . . 123
6.2.1 Self-Referential Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.2.2 Introducing the Notion of Theory Absorption . . . . . . . . . . . . . 125
6.2.3 Determinants of Theory Absorption . . . . . . . . . . . . . . . . . . . . . 127
6.2.4 Some Theoretical Implications of Theory Absorption . . . . . . 128
6.3 Theory Absorption among Bounded Rational Decision-Makers . . . . 132
6.3.1 Individual Absorbability of Theories . . . . . . . . . . . . . . . . . . . . 133
6.3.2 Full Absorbability of Theories . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.3.3 Partial Absorbability of Theories . . . . . . . . . . . . . . . . . . . . . . . 138
6.4 Applications of Theory Absorption to Economic Policy
Advising . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7 On the Absorbability of Economic Theories – An Experimental

Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.1 The Experimental Method in Economics . . . . . . . . . . . . . . . . . . . . . . . 146
7.2 A Possible Experimental Approach for Testing
the Self-Referentiality and Absorbability
of Economic Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.3 Related Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.4 Some Preparatory Attempts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
7.4.1 An Experimental Attempt of Debiasing
the Conjunction-Effect Bias through Meta-Information . . . . . 152
7.4.2 A Classroom Experiment on Theory Absorption
in Multilateral Integrative Negotiations . . . . . . . . . . . . . . . . . . 153
7.4.3 An Experimental Guessing-Game in the Classroom
with Information Feed-Back and Meta-Instructions . . . . . . . . 154
7.5 On the Absorbability of Guessing Game Theory . . . . . . . . . . . . . . . . . 155
7.5.1 On the Guessing Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.5.2 Iterated Elimination of Dominated Strategies
in Guessing Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.5.3 The Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
7.5.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
7.5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.6 An Experimental Study on the Absorbability of Herd Behaviour
and Informational Cascades Theories . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7.6.1 Herding and Informational Cascades . . . . . . . . . . . . . . . . . . . . 171
7.6.2 A Simple Model of Informational Cascades: A Dichotomy
Choice Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
7.6.3 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
7.6.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
7.6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.7 Concluding Remarks on the Experimental Examination . . . . . . . . . . . 182
xii Contents
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
1 Instructions to the Experiment on the Absorbability of Guessing
Game Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
1.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
1.2 Additional Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
2 Instructions to the Experiment on the Absorbability
of Informational Cascades’ Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
2.1 Additional Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Introduction
Since the individuals are not just stimulus-response machines but more complex
beings that think and are simultaneously conscious of their thought, reflexivity is
potentially involved in all human acts of cognition and in all conceptualizations.
On this basis, each human discourse can be characterized as a way of thought for-
mulation and therefore, reveals a self-referring nature. On this level of reflexivity,
the individual thought shapes beliefs and mental representations which give life to
mental models and strive to predict future events and developments to support the
individuals in their decision-making. Such mental models are reflected by the in-
dividuals themselves and on the situation they are confronted with. According to
the result of this recursive application, the individuals will then decide which model
they want to refer to, or in other words, which model they want to absorb.
Similarly, the individuals can make use of social theories and predictions which
can therefore yield recursive effects and interfere with the phenomena they aim to
depict. Revealed theories, if accepted, may influence the behaviour or the agents
they focus on, either in the sense of validation of the theoretical content or in that of
its rejection.
This dissertation tries to discuss the implications of the recursive or self-reflexive
effects of economic theories on bounded rational economic behaviour and inter-
action. The mechanisms through which bounded rational actors perceive the self-
referential nature of economic theories and might absorb their prescriptions will be
focussed and deepened both from a theoretical and an experimental point of view,
according to the evidence of two experimental studies.
First of all, the polymorphism of reflexivity and its involvement in any form of
human understanding, activity and conceptualizing will be underlined, and the con-
cept of “self-reference” will be defined on basis of different kinds of reference rela-
tions. Some common varieties and possible taxonomies for self-reference will then
be discussed and some implications for formal and natural language presented. In
order to test self-reference at logical consistency and to extrapolate some guidelines
for its legitimacy, its relation with paradoxes will be deepened. Some considera-
tions on the role of self-reference for human understanding as well as for social and
individual decision-making conclude the first chapter.
S. Sandri, Reflexivity in Economics: An Experimental Examination 1
on the Self-Referentiality of Economic Theories,
c Springer-Verlag Berlin Heidelberg 2009
2 Introduction
The investigation of the notion of “self-reference” will, in particular, indicate that

neither social reality, nor its observation or description can be disentangled from
their self-referential character, which means that reflexivity and its implications
should be of central concern for the social sciences and research. In particular, it can
be distinguished from two orders of reflexivity affecting social reality: a first order
of reflexivity which involves social reality per se and consists of social phenom-
ena which may recursively have an effect on themselves, and a second order which
concerns the “discourses” on social reality, such as social sciences and theorizing.
The analysis of the first order of reflexivity will be explored in Chap. 2, which
offers an overview of common reflexive social phenomena ranging from anthropol-
ogy, linguistics, law, politics, sociology and psychology. The chapter concludes with
examples of reflexivity which involves economic reality.
In order to discuss the implications of reflexivity for the social sciences and the
predictability of social and economic reality, Chap. 3 adopts a constructivist per-
spective which considers reflexivity as an inescapable basis of all that can be thought
and conceptualized and as a key concept for deepening human cognition and behav-
iour. On this basis the implications of reflexivity of the social sciences for the pre-
dictability of social and, in particular, economic reality will be widened because they
emerge from the comparison between the processes of explaining and predicting so-
cial reality. The self-altering, reflexive effect of social predictions which emerges in
this insight will be discussed as well.
Since the recursive effects of economic theories and predictions on the dynamics
of an economic system essentially depend on their understanding and acceptance by
the economic actors, the discrepancy between the neoclassical rationality standard
and the observable cognitive limitations to the subjective rationality will be explic-
itly analyzed. Therefore, Chap. 4 focuses on the rationality debate in economics and
discusses the unsolved dualism between rationality assumption and psychology of
choice. The role of heuristics in orienting bounded rational decision-making will be
investigated in Chap. 5 and presented together with some elements of the debiasing
research which can serve as a design of a suitable framework for the experimental
analysis of the self-referentiality of economic theories and their absorption.
Chapter 6 specifically addresses the concept of “theory absorption” which con-
stitutes a tool for analysing the possible causal role of theories on bounded rational
economic behaviour and for testing theoretical frameworks at their descriptive and
normative validity. This concept will be discussed both in the spirit of the neoclas-
sical theory and of the bounded rational analysis.
This analysis will, in particular, indicate that the recursivity of economic theories
and their absorption differ from case to case, which means that the inquiry of this
topic needs to be supported by empirical and experimental findings. A possible ap-
proach to the experimental analysis of self-referentiality and absorption of economic
theories will thus be sketched in Chap. 7, which mainly focuses on the results of the
two experimental studies of Morone, Sandri, and Uske (2008) and Fiore, Morone,
and Sandri (2007).
The first experiment examines the absorbability of equilibrium predictions on
guessing games, whereas the second discusses the absorbability of the informational
Introduction 3
cascades’ theory. Both settings have several attractive characteristics for testing the-
ory absorption among bounded rational decision-makers. They namely permit the
effects of rationality to be disentangled from social preferences and have at the same
time a very simple economic interpretation. Further, both the guessing game and
herding behaviour represent interactive settings in which the individuals, in order
to achieve a satisficing result, have to anticipate the others’ behaviour and in which
individually optimal behaving does not per se ensure success.
In particular, the first experiment discusses aspects of full and partial theory ab-
sorption in repeated one-shot p-guessing games with changing parameterizations,
while the second experiment argues whether providing individuals with theoretical
information on informational cascades affects the overall probability of herding phe-
nomena to occur. The second experiment also argues whether an incorrect cascade
can be reversed because of bounded rational adapting to the theory’s prescriptive.
Chapter 1
Reflexivity and Self-Reference
The phenomenon of referring is pervasive and regards all fields of human thought
and activity, so much that it appears to be an inescapable basis of all that can be
thought, conceptualized and expressed.1 The human capability of referring creates
the basis for ordering the subjective perception of the world, for interpreting events,
for interacting with others, etc., thus creating the basis for all activities which re-
gard human cognition and which are essential for individual survival. Being able
to establish self-references is even a necessary prerequisite for self-change and be-
havioural adjustment. Furthermore, the reflexive capacity underlies basic problem-
solving abilities and makes mental adaptiveness possible.2
Consciousness (in the form of self-consciousness) can be identified as the main
source of reflexivity for human thought and action. Individuals think and are si-
multaneously conscious of their thought, so that all discourses are both directed to
outward reality (the external world) and to the inner reality of the individual who
formulates them, since she is conscious of expressing them. Therefore it can be
said that each human discourse, being a human way of thought formulation, has a
self-referring nature.
This chapter is dedicated to the analysis of the polyvalent concept of “self-
reference.” After its definition which will be accompanied by an overview of the dif-
ferent kinds of reference relations some common varieties and possible taxonomies
for self-reference will be presented. The polymorphism of self-reference will be il-
lustrated by its implications for formal and natural language. Logical consistency
of self-reference in its different forms and contexts of appearance will then be dis-
cussed, in that the relation between self-reference and paradoxes will be deepened
and some guidelines for testing the legitimacy of self-references will be extrapo-
lated. The chapter concludes discussing the role of self-reference for human under-
standing as well as for social and individual decision making.
1 Cf. Bartlett (1987, p. 5).

2 Cf. Bartlett (1987, p. 6).
6 1 Reflexivity and Self-Reference
1.1 Definition
Reflexivity is “the quality or state of being reflexive,”3 whereas “reflexive” stems

from the past participle of Latin verb reflectere, i.e. reflexus, which means reflec-
tive, turned back.4 The concept of reflexivity thus applies to something capable of
bending back a beam of light as well as to something that is directed or turned back
upon itself. In this second meaning, reflexivity as reflection upon itself encompasses
mental acts of thought, something capable of reflecting, or something relating an en-
tity to itself.5
The concept of reflexivity applies to the broad class of self-reflections, though to
all “self-x,” whereas “x” can be any possible entity. “‘Reflexivity’ is the generic name
for all kinds and species of circularity. It includes self-reference of signs, the self-
application of principles and predicates, the self-justification and self-refutation of
propositions and inferences, the self-fulfilment and self-falsification of predictions,
the self-creation and self-destruction of logical and legal entities, the augmentation
and self-limitation of powers, circular reasoning, circular causation, cyclic and spi-
ral powers, feedback systems, mutuality, reciprocity, and organic form. It includes
the fallacious, the vicious, the trivial, and the question begging, but also the sound,
the benign, the useful, and the inescapable.”6 So, dealing with reflexivity means to
deal with a heterogeneous class of elements, the only common feature among them
is that they are based on a circular self-referential relation.
The related concept of “self-reference” “is involved in a description which refers
to something that affects, controls or has the power to modify the form or the va-
lidity of that description. [. . .] In this general sense, self-reference establishes a
circularity that may involve not only referential but also causal, interpersonal or
instrumental relations and thereby constitute a unity of its own.”7 Therefore, self-
reference is strongly related to the concept of feedback and based on its mechanism,
whereas “a positive [resp. negative] feedback loop [is] a chain of cause-and-effect
relationships [that] closes on itself, so that increasing any one element in the loop
will start a sequence of changes that will result in the originally changed element
being increased [resp. decreased] even more”8 . An example of a positive feedback
loop is the increase of money in a savings account due to the interest rate.
If the synonymous use of the concepts “self-reference” and “reflexivity” is legit-
imate or not is still a quite debated question. Some9 try to interpret self-reference
in a restrictive sense as a subspecies of reflexivity. As their opponents do, they also
adhere to the view that self-reference occurs whenever an object refers to itself.
However they merely grasp such a definition as denotation or denomination, thus
3 From the Merriam-Webster’s Unabridged Dictionary (2000).

4 Idem.
5 Idem.
6 Suber (1987b, p. 259).
7 Cf. Krippendorff (1986), at http://pespmc1.vub.ac.be/Asc/SELF-REFERE.html.
8 Cf. Bartlett (1987, pp. 21, 22).
9 As pointed out for instance in Scheutz (1995, p. 19).
1.1 Definition 7
as something which aims at depicting or characterising itself. Therefore, they es-

sentially restrict the application of the concept of self-reference to linguistic species
of reflexivity, both regarding natural or formal languages. According to the sup-
porters of this distinction, the concept of “reflexivity” labels the whole categories of
reflexive entities, forms and structures, and should therefore resume all kinds of self-
referential phenomena, self-references included. This distinction sounds somehow
artificial and rather arbitrary, since splitting reflexive phenomena according to the
entity that self-refers seems to be quite a trivial distinction. In order to be reflexive,
a phenomenon needs, to be based on a reflexive relation, which can be of different
kinds. It therefore makes more sense to distinguish among reflexive phenomena just
when they rely on different sorts of reflexive relations and classify them correspond-
ingly. Keeping with Bartlett and Suber’s (1987) tradition, since both the concept of
“recursivity” and that of “self-reference” are similarly based on a feedback loop
mechanism, they will be henceforth used as synonyms.
The first studies on reference concerned mostly linguistic questions, whereas a
self-referential sentence is “a statement that refers to itself or contains its own ref-
erent.”10 Lots of elementary linguistic forms present a self-referential character:
definite descriptions or proper names, for instance.11 There are different degrees of
semantical self-reference, depending on if the sentence refers exclusively to itself,
or also to itself as a member of the whole class of reference. An example of a to-
tally self-referring sentence is “This is a short statement,” while the sentence “All
the sentences on this page are meaningful” is just partially self-referring. Besides,
a statement can also be incidentally self-referring if it can be interpreted as a self-
reference, though only if the statement itself belongs to the sub-class to which it
refers; for example, “Some sentences on this page are meaningful.”12 A statement
may also be self-referential if it refers to itself via another statement. An example
of such an indirect self-reference is if on one side of a blackboard is written “The
statement on the other side of the blackboard is true,” and on the other side is written
“The statement on the other side of the blackboard is false.”
Reflexivity has often been interpreted13 as a possible menace to logical reason-
ing and as potentially leading to paradoxes. Since there are different forms of self-
references, it is not possible to draw general conclusions on the logical legitimacy
of reflexive mechanisms. An attempt in this direction will be done at Sect. 1.7 of
Chap. 1. The famous Liar’s paradox, “p is false,” leads to a contradiction, whether
p is false or true, and therefore it constitutes an example of a misleading, malignant,
self-reference. On the other side, the sentence “q is true” cannot be anything but
true and valid, since the opposite (the Liar’s paradox) is contradictory per se. Thus,
this is the case of a harmless, benign, self-reference. Generalizing those examples,
self-references that present a self-reinforcing character - thus relying on a positive
10 Cf. Krippendorff (1986), at http://pespmc1.vub.ac.be/Asc/SELF-REFERE.html.
11 Cf. Bartlett (1987, p. 5). Going more into details, there is a vivid debate on the legitimacy of
self-referential sentences, as well as on the relations between linguistic self-referential sentences
and paradoxes. See for more information Bartlett (1987) and Whewell (1987).
12 Cf. Whewell (1987, pp. 32, 33).
13 This relates particularly to the fields of philosophy, logic and scientific methodology.
feedback loop - have a stabilizing effect on the phenomena they concern, while re-
flexivities acting in a self-refuting way - i.e. based on a negative feedback loop -
have a destabilizing nature.14
1.2 Reference Relations
A reference relation R connects a class of referring objects with a class of objects

being referred to, and can be expressed on set-theoretical terms as:
(a, b) ∈ R iff b is referred to by a
The set of a’s for which there is a b-(a, b) belonging to R - constitutes the domain
of R, which can be denoted as dom(R). While the set of b’s for which there is an
a-(a, b) belonging to R - is the range of R, denoted by ran(R).15 As shown at Fig. 1.1,
the reference relation R can be illustrated by dom(R) ∪ ran(R).
Obviously, if domain and range of R are strictly separated from each other, i.e.
if dom(R) ∩ ran(R) = Ø, no self-reference can occur. A self-referential relation re-
quires dom(R) ∩ ran(R) = Ø, otherwise there is no possibility for an element of
the domain to refer to itself as an element of the range. Only if this prerequisite is
fulfilled, a self-reference may establish.
Self-reference always articulates on circular basis, whereas it can be distin-
guished between direct and indirect self-reference respectively if there is a loop
at a or if a is contained in a cycle.16
Such a difference can be illustrated, considering sentence T , “This sentence is
true,” as an example of direct self-reference. Sentence T relies on three references:
first, it refers to the sentence itself by means of the term “this”; second, it refers to
the “is”-relation; and third, to the concept of truth. The sentence and its reference
relations can be visualized as in Fig. 1.2.
dom(R)
a2 a4
a1 a3
ran(R)
b1 b2 b3
Fig. 1.1 Reference relation R
(author’s representation)
14 Cf. Davis and Klaes (2003, p. 333).

15 Cf. for definitions and notation Bolander (2002, p. 9 ff).
16 Cf. Bolander (2002, pp. 12–14).
1.2 Reference Relations 9
Fig. 1.2 Reference relations

for “this sentence is truth”
(Bolander, 2002)
ran(R)
dom(R)
S1 S2
“is”-relation
truth
Fig. 1.3 Reference relations for S1 and S2 (Bolander, 2002)
The loop at T hints at a direct self-reference, because it means that T itself refers
to the sentence T , so that there is a self-referential relation (T, T ) ∈ R.
On the other hand, an indirect self-reference establishes when two separated en-
tities refer to each other. Consider, for instance, the following sentences:
S1 : The sentence S2 is true S2 : The sentence S1 is true
Neither of these sentences is directly self-referential. Nonetheless, since dom(R)
∩ran(R) = Ø, the prerequisite for a self-reference to occur is given considering
domains and ranges of both sentences. In addition, elements of the domain (a) are
contained in a cycle in the reference relation where, as in Fig. 1.3, S1 refers to S2
and S2 refers again to S1 , so that an indirect self-reference establishes.
Self-reference is particularly likely to take place in situations involving reflec-
tion, universality and ungroundedness.17 Each of those properties, either occurring
together or singularly, might account for a self-reference to establish as well as for
its specific features.
17 Cf. Bolander (2002, p. 10 ff).

1.2.1 Reflection
The notion of “reflection” has already been used in defining the concept of “self-
reference,” reflection meaning “bending back” and self-reference consisting of
bending back to itself. Whenever something can be viewed back as in a mirror,
it is commonly spoken of as reflection. In the eyes of the reproduced object, its
mirroring represents an outward perspective on itself.
Thus, a reference relation R can be said to have a “reflection,” if and only if for
all elements involved in R there is a referring object (formally dom(R) ⊆ ran(R)).18
In other words, for every element r of the domain dom(R) the relation R associates
an element q belonging to the range ran(R).
Self-reference is just a special case of reflection, which happens whenever re-
flection links equivalent elements of range and domain with each other (i.e. q = r).
As will be illustrated in the following paragraph, this always occurs when reflection
combines with universality.
1.2.2 Universality
When statements regard the totality of entities they are discussing, they can be said
to be universal. In other words, universal statements encompass a totality of ele-
ments, and formally, an object a belonging to the domain dom(R) is called universal
if (a, b) ∈ R for every b ∈ ran(R).19
A universal object a implies a self-reference whenever it belongs to the domain
of a relation R that has a reflection (in the above specified sense).20
This result can be derived as follows: assume a to belong to dom(R) being uni-
versal, as for instance the utterance “all sentences,” and R to have a reflection,
so that dom(R) ⊆ ran(R). Because of that, and since (a, b) ∈ R for every b ∈
ran(R), (a, a) ∈ R, which is a self-reference, holds.
When reflection combines with universality creating a self-reference, paradoxes
are most likely to appear. An example of this is Cantor’s paradox, which proves an
inconsistency in set theory, stating that there is no greatest cardinal number. Cardinal
numbers as such must always admit for an ordering, while Cantor’s paradox involves
the universal concept of infinite sizes, which should be infinite per se, so that no
greatest cardinal number can be determined.21
According to Whitehead and Russell (1910) the responsibility for similar logical
inconsistency should be ascribed to universality rather than to self-reference. This
18 Cf. Bolander (2002, p. 18).

19 Cf. Bolander (2002, p. 11).
20 Idem.
21 Assuming C to be the largest cardinal number. Its power set, i.e. the set of all subsets of the set C,
is 2C , which has per definition a strictly larger cardinality than C. But this contradicts the premise,
C being the largest cardinal number, from what follows, the largest cardinal number cannot exist.
1.3 Varieties of Self-Reference 11
is because, as stated by the theory of types, “it is impossible or meaningless to state

propositions which have an unrestricted possible range of values, or which, in any
sense, are arguments to themselves.”22
1.2.3 Ungroundedness
Finally, self-reference is likely to appear in situations that show an ungrounded na-

ture, i.e. in situations that rely on an infinite regress. An object belonging to the
domain of a reference relation (a ∈ dom(R)) can be said to be ungrounded, if there
is an infinite path in the graph of the reference relation R beginning from a. Oth-
erwise, such path being finite, a is grounded.23 Obviously, for ungroundedness to
occur there must be some common elements between the range and the domain of
a reference relation, i.e. the intersection between dom(R) and ran(R) should not be
empty. Again, this is the same condition self-reference requires.
However, even if self-reference always leads to ungroundedness, ungrounded-
ness is not necessarily accompanied by self-reference. It can be thought as a conse-
quence of self-reference, since any self-referential object establishes a cycle, which
inevitably results in an infinite regress, thus an infinite path passing through the
entities involved in the self-referential relation.
Ungroundedness, either associated with self-reference or not, often links with
paradoxes. Consider for example Yablo’s paradox that posits, for an infinite se-
quence of sentences Si , “All sentences S j with j > i are false, i, j = 1, . . . ∞.”24
Thus, when it comes to identifying possible sources of paradoxes, ungrounded-
ness should be also considered.
1.3 Varieties of Self-Reference
It is possible to enumerate many types of self-references and several classifications

of this concept have been proposed. In order to classify reflexivity, two different
approaches can be essentially followed: it can be focused either on the entity in-
volved in the self-referential relation, or on the kind of self-referring relation. Be-
cause of the pervasiveness of reflexivity, the first method may not offer manageable
taxonomies, while the second may suffer from a lack of comprehensiveness.
In the spirit of the first approach this paragraph concentrates on a variety of self-
referring entities. In offering an overview of some well-known forms of reflexiv-
ity without the claim of being comprehensive, pervasiveness and polymorphism of
22 Cf. Weiss (1992, p. 37).

23 Cf. Bolander (2002, p. 12).
24 Cf. Bolander (2002, p. 13).
self-reference will be illustrated. Particular attention will be paid to the role reflex-
ivity may assume in different contexts and for the various disciplines.25
The second approach for classifying self-reference differentiates the various
forms of reflexivity according to the sort of recursive relation they imply and will
be discussed in the next paragraph (cf. Sect. 1.4 of Chap. 1).
As reflexivity is involved in all discourses that seek to know the presupposi-
tion involved in knowing it applies to an incredible range of phenomena, spreading
from the symbolic attribution of meaning to language codification and its acceptance
leading to discourses referring to other discourses, for instance science of science,
i.e. scientific methodology. According to Lorenzen26 the acceptance of elementary
sentences (thus of natural language) constitutes the condition of any discourse and
has a recursive character per se. The symbolic process, which originates from both
natural and formal language is a self-referring system, because it puts the basis for
its own regulation. Symbols testify moreover self-awareness and self-consciousness
of individuals: they were created to express human thoughts, so that individuals are
at the same time users - i.e. subjects-and content - i.e. object - of symbolism.27
The first studies on reflexivity concerned linguistics and focused on both seman-
tic reflexivity and self-referential elements of the natural language.
Natural language always grants for possibilities of referring to semantic concepts
through linking language to a class to which language itself can refer, typically to
its own content. Semantical self-references in particular may establish a combina-
tion of concepts truth and falsity since they establish a semantical link per se, and
because they articulate on a self-referring basis.28 Such a link can then become
self-referential, if it relates a statement or its semantical content to itself, e.g. in the
sentence “This is a true statement.”
That natural language allows for semantical reflexivities, however does not add
much to the legitimacy or congruence of semantical referring. A vivid debate has
flourished about this topic,29 which has become even more complicated, considering
that there are different degrees of semantical self-reference as well as different ways
a statement may semantically refer to itself. While aspects of semantical reflexivity,
as well as some conditions for its logical consistency, will be more closely discussed
in Sect. 1.5 of Chap. 1, it should be mentioned here that semantical self-reference
may either be based on self-validating (respective self-refuting) dynamics, or rely on
meta-logical linkages, for instance in meta-statements. Those are statements whose
content applies to the content itself, or in other words, statements speaking about
themselves, e.g. “This is not a statement about Socrates.”30 The first are statements
involving judgement of themselves as expressed by sentences involving truth, fal-
sity, plausibility etc.
25 The main reference for this paragraph is Bartlett (1987).

26 As mentioned by Bartlett (1987).
27 Cf. Winrich (1984, p. 988).
28 Cf. Winrich (1984, p. 988).
29 For a contribution, see e.g. Whewell (1987).
30 The example stems from Whewell (1987, p. 35).
1.3 Varieties of Self-Reference 13
When a sentence is applied for its own legitimization or defence, the threat of log-
ical rudeness is always present. This occurs in ad hominem argumentations which
cannot be falsified and are therefore irremediably sterile. Although the menace of
logical rudeness can be mainly related to self-validating semantical reflexivity, it
should be contemplated even for dealing when with meta-statements.
Self-referential elements are present in all natural languages, and they establish
a circular recursive semantical relation and therefore they rely on similar character-
istics. Several attempts have been made to reduce them to a common denominator.
Peirce31 refers to these self-referential elements as “indexical signs,” Russell32 as
“egocentric particulars” and Reichenbach33 as “token-reflexive words.” All these
elements refer to the argument of a sentence, whereas the subject, object, or con-
ditions stated can be hereby meant. Self-referential elements are for example terms
as “I,” “here,” “now” etc. These terms all share the property of referring, in a way
that is relative to the speaker who uses them. While Peirce’s analysis follows an
enumerative approach, Russelll and Reichenbach strive toward reducing all such
expressions to the utterance “this.” Furthermore, it can be noted that recursive re-
lations based on self-referential elements of discourse typically do not create the
conditions for paradoxes to take place.
A further variety of reflexivity can be generated by dictionary reference. This
occurs most in an indirect way, whenever a circular relation between two or more
references is established. A dictionary provides a definition for each word, which
cannot be expressed but in terms of other words. Thus, if a concept has been ex-
plained recurring to a second one, the latter might be of a certain likelihood used for
defining the previous concept. In this case a direct self-reference would occur. To
a greater probability, however, self-reference would establish indirectly, i.e. after a
long chain of words linked together by mean of reference relations: a first concept is
defined through a second concept, which is again explained through a third, a forth,
a fifth concept and so on, up to when the first concept will be employed again. In
this way an indirect self-reference would take place creating a cycle that links the
concepts altogether in an ungrounded regress. Dictionary reference is not merely a
speculation or a possibility, but something inevitable since languages do not possess
but a finite number of words, so that any dictionary must contain words which are
indirectly defined in terms of themselves.
As pointed out by Wittgenstein (1999), dictionary reference also articulates at the
semantic level, mining with ungroundedness the entire semantic construction of lan-
guage. A dictionary relates words to each other, so that without knowing any words
it is not possible to learn anything from it. Thus, a dictionary constitutes a closed
self-validating system and suffers from a logical point of view from unground-
edness. Ostensive definitions,34 which convey the meaning of a term by pointing
out the object itself or examples for the word to be explained would represent
31 Cf. Peirce (1931–58).

32 Cf. Russell (1940).
33 Cf. Reichenbach (1947).
34 Cf. Wittgenstein (1999).
a possibility to escape the self-referentiality of defining and of dictionary attribu-

tion of meanings. Their applicability however remains confined to pure speculative
purposes and does not make much sense for practical usages.
Reflexivity plays a particularly significant role both for philosophical argumenta-
tion and for phenomenological analysis. “Petitio principii,” “reductio ad absurdum,”
and “ad hominem arguments” are just some examples of circular figures that are of-
ten used in philosophy. Philosophy relies much on reflexive approaches in particular
when it comes to arguments with transcending as well as transcendent orientations.
These are respective arguments trying to delimit the realm of possibilities and argu-
ments establishing a linkage between entities belonging to different ontological or
logic levels.
The solution for such argumentations lays in many cases on self-application of
principles or categories, self-validation (respective-refutation) of theories or their
propositions, as well as on the self-supporting nature of inductive speculations. Re-
flexivity supporting philosophical argumentation operates at a procedural level and
regards the form of the argumentation while reflexivity that may be involved in phe-
nomenological elaborations concerns the content of the argumentation itself. E.g. in
Husserl’s phenomenology,35 phenomenology can be reduced to a science of science,
thus to a theory of theories. The radical constructivist interpretation of any phenom-
ena as a construct of an observer’s mind posits the recursive character of ontology
and phenomenology, because it ascribes to an autopoietic origin of the mind.
Self-referential techniques have been exemplarily applied in proof theoretical ap-
proaches to mathematics and their analysis leads to un-reassuring conclusions since
it indicates the paradox nature of mathematics. This is because, self-references are
typically involved in mathematical reasoning when it comes to extend the analy-
sis to the borders with infinite (thus universal) entities. Concepts such as “infinity,”
“non-closed system,” “incompleteness,” “non-excludability,” etc. are most likely to
create the basis for paradoxes arising on a self-referring basis since they involve
both infinity and negation. The scale of the problem can be appreciated thinking
that as soon as self-reference is ruled out, mathematics (that is in the end a formal
language and as such a closed self-referring system) becomes unworkable. Corol-
laries of that nature are provided e.g. by Gödel’s numbering,36 and by Russelll’s
paradox developments of the axiomatic set theory.37
Other self-referring applications of mathematics can be found in the computabil-
ity theory and are e.g. illustrated by the reflexive functions theory, as well as by
research on artificial intelligence. Studies of artificial intelligence deepen the no-
tions of self-correcting, -regulating, -organizing and -reproducing systems, of self-
initiated learning, etc.38
A way to codify the perception of space and time when it comes to understanding
infinite phenomena is to use recursive signs and representations.39 The closed loop,
35 Cf. Elveton (2000).
36 Cf. e.g. Smullyan (1991).
37 Cf. e.g. Winrich (1984).
38 For an overview on artificial intelligence, see e.g. Luger (1995).
39 More on that is presented e.g. in Priest (1987).
1.4 Taxonomies of Self-References 15
the Moebius strip, the Klein bottle and the Rimannian model of a closed universe are
example of recursive structures which are used for spatial analysis. Further, closed
temporal loops can constitute self-referring tools for capturing temporal cyclic dy-
namics and can be applied to the most various fields of analysis, that range from
particle physics to the analysis of economic cycles.
Physics may be affected by reflexivity as well as illustrated by the puzzling re-
sults in quantum mechanics and general relativity. Despite traditional view of a nat-
ural scientist as perfectly disentangled from the reality she observes, experiments in
quantum mechanics reveal that in some situations measuring system, observer and
quantum phenomena to be measured “constitute a system which itself reflexively
defines properties of the phenomena which may be measured.”40 Functionally inter-
dependent descriptions represent a reflexive structure which is often used in physics
for example in general relativity, density and gravitation are defined as functions
of the space curvature, in models of closed universe, which is at the same time un-
bounded but finite.
Up to this point, all reflexivities mentioned have in common that they are to be
more or less the product of human thought and activity. This is concurrent with
the thesis interpreting self-consciousness as main source of reflexivity. It should be
however noted that reflexivity may as well exist independent of human activity, as at-
tested by numerous reflexive natural phenomena. For instance, recursivity is implied
in viral reproduction, which can be interpreted as a self-replicating process. A fur-
ther example is offered by the biological notion of the self-organizing system.41
1.4 Taxonomies of Self-References
Basically, it can be distinguished between self-references which establish circular

(i.e. closed) or processual (i.e. open) relations. While incompleteness might affect
the legitimacy of closed self-references, processual open self-references are mostly
immune to such a threat.
Self-referential relations can establish tautological, set-theoretical, pragmatical
or metalogical relations.42 A tautological self-reference can be characterised as a
static relation, or in other words as a self-reference that does not add anything but
redundant information to its predicate. A set-theoretical reflexive relation generates
mostly paradoxes and it appears when set-memberships are used in a reflexive way.
Groucho Marx’s gag: “I don’t want to belong to any club that will accept me as
a member,” is an example of such reflexivity. When the content of a sentence and
the sentence itself refer to each other, as in the sentence “there are no truths,” it
can be spoken of as a pragmatic or performative self-reference. Finally, the relation
“between a truth-functional referring proposition and the set of conditions which
40 Bartlett (1987, p. 13).
41 This notion can be applied to many phenomena like e.g. the spontaneous folding of proteins,
formation of lipid membranes and homeostasis. For more, see e.g Solé and Bascompte (2006).
42 This classification of self-references is freely based on Bartlett (1987).
are necessary in order for the proposition to be capable of referring at all”43 can be
defined as meta-logical. Human understanding in general and scientific reasoning
often involves this sort of referring.44
Another interesting taxonomy of self-references has been proposed by Davis and
Klaes (2003). They distinguish between immanent, epistemic and transcendent re-
flexivity depending on which levels the reflexive relation involves. Immanent reflex-
ivity means a reflection from an entity to itself, while epistemic reflexivity results
from a conscious act of a subject referring to itself. Eventually the transcendent
reflexivity almost coincides with the meta-logical self-reference presented above.
Scheutz (1995) restricts self-reference to linguistic forms of it and interprets re-
flexivity as the general class to which it belongs.45 On that basis he distinguishes
between reflexivity of symbols, syntactic and semantic reflexivity, set theoretical,
pragmatical or performative, and metalogical or transcendental reflexivity.46
Even if it seems impossible to reduce the different way of classifying self-
reference to a unique resuming taxonomy, an attempt can be made in this direction,
since it can be noted that there is in fact a common denominator in analysing reflex-
ivity, and that there are some categories that similarly appear in the different classifi-
cations. In particular, there is a certain consensus in recognizing the specificity of the
mechanisms involved by semantical, pragmatical, metalogical and computational -
in the sense of set theoretical - reflexivity.
Resuming, what emerges from the analysis of the concepts of “self-reference”
and “reflexivity” is their occurrence in uncountable forms and their application to
a broad spectrum of phenomena. After this introductory overview on varieties and
general taxonomies the concept of self-reference will be related to those aspects that
are particularly relevant for human conceptualizing, namely reflexivity in natural
and formalized language. The first will be explained in the next paragraph (Sect. 1.5
of Chap. 1), where some of the possible ad hoc classifications will be presented
and linguistic terms and general mechanisms through which self-reference may
establish will be discussed. The latter form of reflexivity is discussed in Sect. 1.6
of Chap. 1, and deals with how self-reference can be expressed in formalised lan-
guages and with the consequences for the consistency of the system to which the
formal language refers.
1.5 Self-Reference in Natural Language
In examining self-reference in grammar, Van Fraassen47 distinguishes between two

kinds of self-referring terms, namely context-dependent and -independent. The
43 Cf. Bartlett (1987, p. 10).

44 More on Sect.1.8 of Chap. 1.
45 This has already been discussed in Sect. 1.1 of Chap. 1.
46 For more, see Scheutz (1995, p. 14 ff).
47 It is here mostly referred to Van Fraassen (1970).
1.5 Self-Reference in Natural Language 17
difference among them lies in having a variable (i.e. “context dependent”), or an

invariable (i.e. “fixed”) addressee. Formulated otherwise, it can be spoken of a
context-dependent self-referring term, when “The extension of the term (or rather,
the extension of an occurrence of that term) depends on the context in which it
occurs.”48
Each type of self-referring term implies that a different mechanism of self-
reference will occur, in order to be distinguished between accidental or func-
tional self-reference: “Whenever self-reference consists in reference to a sentence
by means of a term occurring in that sentence [. . .], and this term is such that its oc-
currences have the same extension, we shall speak of accidental self-reference. [. . .]
When self-reference is produced through the use of terms with context-dependent
reference, we shall speak of functional self-reference.”49 This taxonomy however
has some shortcomings in that it exclusively sticks to the dimension of context-
dependency without capturing other variables that can be made responsible of differ-
ent ways for self-referring. For example, Van Fraassen’s taxonomy does not enhance
to distinguish between total and partial, between direct and indirect, or between be-
nign and malign referring.
According to this insight, the classification of self-reference in natural language
proposed by Martin50 goes beyond Van Fraseen’s, in that it contemplates two levels
for evaluating self-reference: at the first level, the difference between benign and
malign self-references is considered, while at the second self-references are classi-
fied in direct, indirect, empirical and general.
Martin belongs to the philosophical tradition interpreting self-reference as a con-
stitutive part of human language and discourses so that its benignity or malignity
may affect logical reasoning. A self-reference can be clearly stated as being ma-
lign, when contradiction follows, whereas it can be denoted as benign if it “can be
judged, in logically unproblematic way, to be straightforwardly true (or false)”.51
The state according to which all self-referring statements cannot be but meaningless
and necessary leading to paradoxes, was firstly advanced by Russell and supported
by his theory of types.52
The second distinction which Martin introduces is centred on the difference be-
tween “self-referential sentence type” and “self-referential token” in the natural lan-
guage:53 “a sentence type is self-referential if and only if every token of that type is
self-referential, and [. . .] a sentence token is self-referential if and only if what one
mentions with the subject expression of the sentence token is the sentence itself (type
or token) which is being used ”.54
48 Cf. Van Fraassen (1970, p. 696).

49 Cf. Van Fraassen (1970, p. 696).
50 See Martin (1967, 1992).
51 Cf. Martin (1992, p. 119).
52 Cf. Russell (1903).
53 Such distinction exclusively applies to natural language and cannot be extended to formalized
language.
54 Cf. Martin (1992, p. 133).
Whenever a sentence type is self-referring, i.e. whenever each of its token is

in the sense of the previous definition self-referential, it can be spoken of direct
self-reference. This is defined in Martin’s words the “crux version” of self-reference.
An example of a reference of this kind is the sentence: “This very sentence is so and
so,” since “the expression ‘this very sentence’ refers to the whole expression of which
it is a part.”55
Direct self-reference differs from the other categories, namely indirect, empirical,
and general reference. These differ in that the first requires that its entire tokens
token are self-referential, while the others include sentences only tokens of which
are self-referential.
The category “indirect self-reference” always applies to a plurality of sentences,
i.e. up to a pairs, as “The next sentence is so and so. The previous sentence is this
and that,” or as Jones saying “Most of Nixon’s assertions about Watergate are false,”
Nixon replying: “Everything Jones said about Watergate is true.”56 Indirect self-
reference, neither sentence being per se self-referential, is a sub-case of reference
to itself by mean of other sentences. Indirect reference does not necessarily imply
indirect self-reference, except when dom(R) ∩ ran(R) = Ø.57 “The next sentence
is funny. The previous sentence is English”58 constitutes an example of indirect
referring sentences that do not imply a self-reference.
The sentence “Sentence (1) is so and so” furnishes an example of empirical self-
reference, which is in the end a sort of hidden direct self-reference.59 Recursivity
occurs as the context, in which a sentence token is inserted, makes the token refer to
the sentence, so that the sentence assumes the character of a self-referring statement.
This is the case of a context dependent (in this sense “hidden”) direct self-reference,
since reflexivity is given by the context, which makes out of (at least) one token a
reflexive term, which directly refers to the sentence.
General referring is applicable when a sentence occurs to be self-referential, be-
cause of referring to a whole class and being in the meanwhile member of it.
Because context dependency matters in Martin’s classification, Van Fraassen’s
and Martin’s taxonomies can be integrated, the latter classification represents
an extension of the first. Van Fraassen’s category of accidental self-reference
applies to Martin’s direct self-reference, whereas under the label “functional self-
reference” all remaining varieties of self-reference, i.e. indirect, empirical and
general self-reference can be subsumed. The distinction of which Martin’s taxon-
omy is based – i.e. the difference between self-references caused by sentence types
or tokens - seems to provide a good perspective for capturing the specificity of the
different mechanisms self-reference in natural language involves. Unfortunately,
Martin’s classification cannot be applied to self-reference in formalized language
because of the impossibility of formally finding an equivalent to the type-token
55 Cf. Martin (1992, p. 133).

56 This example refers to Kripke (1975).
57 Cf. Sect. 1.2 of Chap. 1.
58 Freely translated from Scheutz’s examples. Scheutz (1995, p. 49).
59 Cf. Scheutz (1995, p. 52).
1.5 Self-Reference in Natural Language 19
distinction. The same applies to Van Fraassen’s classification, as well. Again, the
appealing possibility of classifying different varieties of self-reference adopting
a unique pattern of classification is precluded because of the polymorphism of
reflexivity.
In his article “Self-Reference and Meaning in a Natural Language”60 Whawell
distinguishes between benign and malign self-reference and is concerned with find-
ing criteria for distinguishing what he calls “legitimate” from “illegitimate” self-
references. He lists three possible ways for a sentence to be self-referential that are
to be totally, partially or incidentally self-referring. In a more general form this dis-
tinction has already been mentioned at p. 3. It will be here more specific related to
the context of natural language and will be then compared and integrated with the
other taxonomies that have been proposed in this paragraph.
Once again Whewell’s classification is based on the different degrees according
to which semantical self-reference can occur: a sentence can either refer exclusively
to itself, being thus totally self-referring, or also to itself as a member of the whole
class of reference. For this to occur a partially self-referring character it should be
assumed. In particular, “A statement is totally self-referring if it refers explicitly to
itself by means of a singular referring expression,”61 i.e. by mean of a term referring
exclusively to itself. This allows for other terms of the totally self-referring sentence
to be other-referring, as for instance in the sentence: “this statement and the previous
one are false.”62 This is still the case of a totally self-referring sentence as the other-
referring term does not modify the totally referring character of the sentence.
Whewell’s classification focuses exclusively on semantic degree of reference,
and grants explicitly the possibility of plural referring, but does not explicitly distin-
guish between direct and indirect referring. The category of totally self-referring
sentences is intended to encompass even those cases that in Martin’s taxonomy
would have been interpreted as examples of indirect self-reference. For example
if on one side of a blackboard is written “The statement on the other side of the
blackboard is true,” and on the other side “The statement on the other side of the
blackboard is false.”63
The second category in which Whewell divides self-reference is that of partial
self-reference. In his words, “A sentence is partially self-referring if it is about
a whole class of statements of which it is itself a member. For instance, ‘all the
statements on this page are true’ and the statement ‘every meaningful but non-
tautological statement must be in principle empirically verifiable.’ ”64 For this sort
of reflexivity to occur the context -in a broad sense – plays an important role, be-
cause no referring tokens, terms or expressions are strictly necessary for such a par-
tial reference. The only requirement being that the sentence establishes a reference
to itself via its meaning. Whewell refers to that by stating that it can be sometimes
60 Whewell (1987).
61 Cf. Whewell (1987, p. 32).
62 Idem.
63 The examples have been literally quoted from Whewell (1987, p. 32).
64 Cf. Whewell (1987, p. 33).
necessary “to go outside the statement in order to know whether it is self-referring

or not”65 . Equivalence with Martin’s general referring can be easily inferred.
The third category of self-reference is that of incidentally self-reference. This
could also be interpreted as a sub-species of partial self-reference considering that
a statement can be incidentally self-referring even when its self-referring nature
emerges. This is not only because the statement is a member of its class of refer-
ence, but also because it belongs to the sub-class of members to which it refers,
as well. This can be illustrated by the sentence: “Some sentences on this page are
meaningful.”66 The other classifications previously presented did not provide a sep-
arate category for this sort of reflexivity and consider it equivalent to partial self-
reference. It seems convenient for analytical purposes to rely on such a distinction,
being the self-referring character of incidental self-referring out of doubt, but at the
same time is not very pronounced, as based on second-order set-membership.
Eventually, Mackie (1973) classifies self-reference on two different levels, by
distinguishing between total and partial, and between explicit and implicit self-
references. A total self-reference is the strictest way of referring, because it requires
a referring term or phrase, e.g. “this sentence.” As posited by Whewell, Mackie
considers partial self-reference to origin from set-membership, but does not further
distinguish between a partial and an incidental form of referring. Whenever self-
reference is established because of the presence of a referring singular terminus it
can be spoken of explicit self-reference. On the contrary, if a quantifier or a general
terminus is responsible for a self-referential occurrence, it can be spoken of an im-
plicit self-reference. These two orders of self-referential relations integrate, in that:
“In general, an explicit self-reference will be total, but it need not be: ‘This state-
ment and others. . .’ or ‘This statement and the previous one. . .’ and so on. And in
general an implicit self-reference will be only partial; but it is only contingently that
an implicit self-reference is a self-reference at all [. . .] and what might have been
only a partial self-reference could contingently be total, e.g. ‘Something written on
this page. . .’ where nothing else happens to be written on the page in question.”67
1.6 Self-Reference in Formal Language
A formal language68 (L) is defined by mathematical or machine processable formu-

las. It can be characterised as the set F of finite sequences of elements drawn from
the finite set A of symbols, so that L = {A, F}. A further differentiation that won’t
be discussed here can be drawn between a formalized logistic system (calculus) and
65 Idem.
66 Cf. Whewell (1987, p. 32, 33).
67 Cf. Mackie (1973, p. 286).
68 Cf. Wikipedia at the voice “formal language,” http://en.wikipedia.org/wiki/Formal language,
23th January 2007.

1.7 Logical Consistency of Self-Reference 21
formalized language system (interpreted language).69 This could be of a certain in-

terest for example for assessing and rethinking the linkages between reflexivity and
human thought.
Self-reference in formalized language is a particular delicate topic, because it is
often linked with logical inconsistency. It will be briefly illustrated in the following
how formal languages may allow self-referential devices, how such possibilities
may be responsible for inconsistency and under which conditions consistency might
be regained.
Many formal languages are equipped with the possibility of establishing self-
referential statements. Smullyan’s approach to this concept it to define a cham-
eleonic term σ, which refers to the same formula in which it occurs.70 In other
words, the term σ is provided of a context dependent semantics, so that it can be
viewed as a formalized translation of the term “this (sentence).”
Barwise and Etchemendy71 adopt a similar mechanism for establishing self-
references in formalized language, that consists in the definition of the formal term
“this,” denoting the proposition in which it occurs.
Similar ways of establishing direct self-referring relations require however quite
a sophisticated formalized language, and cannot be applied in elementary languages,
as e.g. in first-order predicate logic. Given the tools of a simple logic, self-references
cannot be established but in an indirect way, and has to involve diagonalization or
equivalent procedures resembling indirect self-reflection in natural language.
The diagonalization method involves a diagonalization function d : N → N de-
fined by d(φ (x)) = φ (φ (x)). If the function d is recursive it then follows that there
exists a formula D(x1 , x2 ) representing d in Q. Given the formula D(x1 , x2 ) the diag-
onalization lemma can be stated as follows: Let γ (x) be a formula of predicate logic
with only x free. Define the formulas α (x1 ) by α (x1 ) = ∀x2 (D(x1 , x2 ) → γ (x2 )) and
β by β = ∀x2 (D(α , x2 )) = α (α ).72
The diagonalization lemma relates to self-reference in that it expresses the same
semantic principle involved in the sentence α : “Sentence α has the property ex-
pressed by β .” As it characterizes itself as being on posses of property β , sentence
α makes a statement about itself.
1.7 Logical Consistency of Self-Reference
The history of reflexivity is tied with that of paradoxes and logics in general, because
the issue of reflexivity has often emerged in combination with the occurrence of
paradoxes.
69 For more on that, see Martin (1992, p. 75), who refers to the contributions of Carnap (1942) and
Church (1951).
70 Cf. Smullyan (1984).
71 Cf. Barwise and Etchemendy (1987).
72 Cf. Bolander (2003, p. 70 ff), who refers to Mendelson (1997).
Paradoxes have always attracted philosophers and logicians, because their analy-
sis is often pushed to the boundaries of human thought and of the validity of logical
constructions. In some cases exploring the insurgence of paradoxes furnished “the
occasion for major reconstruction at the foundation of thought.”73 Reflecting on
ambiguities which underlie paradoxes is not just a sterile speculation as it might
sound at first sight, but a feasible method to achieve scientific advances.
1.7.1 Self-Reference and Paradoxes
The concept of a “paradox” can be applied to statements that appear plausi-

ble and true, but in fact imply something contradictory. A paradox is defined as
“a statement that is actually self-contradictory and hence false even though its true
character is not immediately apparent”74 or “an argument that apparently derives
self-contradictory conclusions by valid deduction from acceptable premises which
defies either intuition or logics.”75 The word paradox stems from the Greek and
combines the preposition “para,” which means “against,”, “beyond” and the sub-
stantive “doxon,” which stands for “thought,” “belief,” “opinion,” so that etymolog-
ical paradoxes refer to something that is “contrary to expectation.” The use of this
concept can be traced back to Plato’s Parmenides where Zeno during a conversation
with Socrates and Parmenides criticises some accepted understandings by means of
commenting on their own basis to paradox conclusions. The most famous of these
paradoxes is the one which says that as long as a tortoise continues to move, it won’t
be overtaken by the fleet-footed Achilles. Though it is now easy to solve Achilles’
paradox through the notion of convergent series, it should not be forgotten that this
notion was introduced in mathematics only in the 19th century. So, what has long
been regarded as an unsolvable incongruence, resulted in the end to be based on a
falsifiable premise.
Paradoxes, whose contradictory conclusion simply stems from a fallacy in their
demonstration, can be called, according to Quine’s classification (1962), “falsidi-
cal.” A falsidical paradox can be in other words solved, and represents no real dan-
ger for logical reasoning. Besides Zeno’s paradox, another well known falsidical
paradox is that of the disproof “1 = 2,” which is based on a wrong division.76
Another class of paradoxes is that of “veridical paradoxes,” which appear to be
absurd, but can be demonstrated to be valid and true. Quine (1962) refers in this
insight to an anecdote about the protagonist of “The Pirates of Penzance,” who after
five birthdays is 21 years old, being born - here the solution is of apparent absurdity -
on February 29th.
73 Cf. Quine (1962, p. 21).
74 Cf. Marriam’s Webster Unabridged Dictionary (2000) at the voice “paradox.”
75 Idem.
76 For x = 1, x = x , so that x − 1 = x − 1. Now, dividing both sides by x − 1 the supposed result
2 2
is x + 1 = 1, from which it follows x = 2, so that 1 = 2 should come out. It is easy to find out the
error in the division of x − 1 by itself, whose result is 0 and not 1. Cf. Quine (1962, p. 22).
1.7 Logical Consistency of Self-Reference 23
There is a possibility that paradoxes may also belong to a third class, which
is in fact the one which “brings on the crises in thought.”77 This is the class of
“antinomies.”
An antinomy is a paradox of a particular kind, which “produces a self-
contradiction by accepted ways of reasoning.”78 That is, it is the acceptance of
the principles stated in the antinomy itself that leads to paradox conclusions.
Antinomies are typically universal statements that, once applied recursively, lead
to an internal contradiction, because they are being the undermined principle in
the rule coincident with the one sustained by the antinomy itself. In this sense, the
paradox nature of antinomies can be ascribed to its application to itself, i.e. in its
self-reference. Antinomies might represent a threat for logical reasoning since they
require a rethinking about the principle stated in the antinomy, which says that it
has either to be abandoned or somehow restricted.
Some examples of well known antinomies will now be presented. From the ex-
ploration of the self-referential mechanisms on which they are based, some tentative
conclusions on how self-reference can affect logical thought will be discussed.
Grelling’s paradox79 deals with the notions of the “autological” and “heterolog-
ical.” The concept “autological” applies to self-describing adjectives, such as the
adjective “short,” which is in fact a short adjective, “polysyllabic,” which is in fact
polysyllabic and so on, so that every autological adjective can be said to be “true
of itself.”80 On the contrary, the category of “heterological” applies to all adjectives
that are not suitable for their own description. The adjective “long,” for example,
does not result to be a long adjective at all, or “tall,” “red,” “monosyllabic” etc.
Such adjectives cannot be applied to their own description, as they are not true of
themselves. Now, what can be said for the adjective “heterological?” It cannot be
autological since this would imply that “heterological” is self-depicting, true of it-
self, which would then require that “heterological” be heterological, but how can
it be true of itself (i.e. autological) if it is not heterological? This chain of reason-
ing cannot be reduced but ad absurdum, leading to what is known as Grelling’s
paradox.
Another famous antinomy is known as the “Liars paradox” and is given by the
Cretan Epimenides saying all Cretans to be liars. To the same category of antinomies
belong the paradox sentences “I am lying” or “This sentence is false.”
Similarly, the exception rule states that there is an exception to every rule. But, if
every rule admits for an exception, what about the exception rule itself? Is there an
exception to the rule saying that there is an exception to every rule? The application
of the rule to itself leads in this case to paradox conclusions, so that the validity of
the rule seems questionable.
Russell’s paradox concerns set-theoretical membership and seriously threatens
the axiomatic fundaments of set theory. It became popularised by the cover story of
77 Cf. Quine (1962, p. 23).
78 Cf. Quine (1962, p. 23).
79 The first formulation of this paradox can be ascribed to Kurt Grelling and Leonard Nelson who
formulated it in 1908. Cf. Quine (1962).

80 Idem.
a village’s barber, who is an adult male who shaves all male villagers who do not
shave themselves or anybody else. The question now is: who shaves the barber? In
set-theoretical terms, Russell’s paradox deals with the question, whether the set of
all sets which do not contain themselves contain itself?
1.7.2 Harmless and Harmful Self-Reference
It has been shown, that contradictions in antinomies are connected with self-
reference. Self-reference is not the sufficient condition to constitute the paradox
nature of an antinomy. Though the paradox nature of antinomies emerges in their
application to themselves, other factors or characteristics must apply to get to para-
dox outcomes. In this sense, antinomies can be extremely useful for spreading some
light on the presumed viciousness of self-reference for logical thought.
A common feature of antinomies is to deal with universal statements, and this
is what makes self-referential devices unavoidable. However, universal statements
can be done without incurring in an antinomy because neither universality nor self-
reference can be made responsible for the paradox arising.
The paradox essence of antinomies can be rather ascribed to the combination
between universality and a negative device of some kind. Grelling’s paradox arises
because the concept “heterological” is only possible as a negative definition, such
as “not-suitable to self-description,” “not-truth of itself,” that is, “not-autological.”
Similarly, the Liars paradox is also produced by the concept of falsity, the negation
of truth. No contradiction would be produced. The Cretan Epimenides said that
all Cretans speak the truth: even if a similar statement has a universal character, it
can be self-reflected without generating any logical contradiction. Moreover, both
Grelling’s and Epimenides’ paradox involve the truth concept, which has been seen
to have a self-referential character in itself, whereas, again, only its negation can be
made responsible for the antinomy. In the same way, the paradox emerging from
the exception rule relies on “exception” being essentially a negative concept, which
means “exclusion or restriction [. . .] by taking out something that would otherwise
be included.”81 Russell’s paradox also inquires negative set-membership.
Such result can be generalized to mean that self-reference can be considered to
be a vicious cycle for logical thought whenever it involves a negative device, thus
whenever it implies a negative backlash. On the contrary, self-reference in conjunc-
tion with positive concepts can be said to be innocuous for logical reasoning. It has
been seen that an essential feature of self-reference is that it is accompanied by infi-
nite regressions as in a feedback-loop. A self-reference will then assume a harmless
or a harmful character, depending on the direction in which the regression works.
In particular, a self-reference which presents a self-reinforcing character, based on
a positive feedback loop, will have a stabilizing dynamic on the entities involved.
81 Cf. Merriam-Webster’s Unabridged Dictionary (2000) at the voice “exception.”

1.8 Reflexivity in Human Understanding 25
On the contrary, reflexivity which acts in a self-refuting way, based on a negative

feedback loop, always works in a destabilizing sense.82
1.8 Reflexivity in Human Understanding
Human understanding is inevitably confined within the bounds of its own conceptual
structure. Such limitations cannot be transcended and inevitably arise in conjunction
with understanding, conceptualizing and information procession of any kind, there-
fore every activity involving intellectual faculties.
To put it another way, this implies that we inevitably reflect the limits of our
understanding on all our intellectual elaborations so that the products of our intel-
lectual operations suffer from the same limitations we do.
In particular, there are two sorts of limitations affecting human understanding
per se, namely pragmatic and metalogical limitations.83 Both can be linked to a
self-referential regress because they are generated by the circular structure which
is established among cognitive faculties that constitute the premises of any under-
standing, as well as among their constructs or elaborations.
Recalling the definitions presented in Sect. 1.4 of Chap. 1, pragmatic or per-
formative self-reference regards the commitments involved in making a certain as-
sumption. It often comes to light in revealing the premises underlying a statement.
A critical use of this kind of self-reference can be done in taking it as a measure of
internal consistency, as there should not be incompatibility between the content of
a theory and the commitments involved in its formulation. Therefore, proving prag-
matic self-reference at its own consistency could be a first test at formal validity of
the statement in which it occurs. Whenever pragmatic self-reference leads to incon-
sistency, a constructive use of the proven inconsistency could be thought to refute
this thesis. Because it is self-referential, logically weak fundaments could be a first
argument in supporting the validity of its negation.
Similarly, human understanding can be critically or constructively formulated
by its meta-logical fundaments, which cannot but develop in a self-referential way.
Meta-logical fundaments constitute the very fundaments of logic, which, as they
stay beyond logics, give it shape and consistency. As said on Sect. 1.4 of Chap. 1,
meta-logical reference occurs in evaluating the set of conditions which make truth-
functional propositions feasible. Meta-logical self-reference can represent a limita-
tion, whenever it implies self-falsifying or self-refusing dynamics, while it can be
interpreted constructively, when it involves self-validation.84
82 Cf. Davis and Klaes (2003, p. 333).

83 Cf. Bartlett (1992, p. 3 ff).
84 For more on the constructive versus critical use of self-reference involving human understanding
see Bartlett (1992, p. 3 ff).

1.9 Self-Reference in Social and Individual Decision-Making
The following paragraph analyses the role of self-reference in social and individual
decision-making and is based on a constructivist approach to human cognition and
decision making85 combined with elements of cognitive psychology.86
The mechanism through which human cognition tries to make the environment
meaningful and predictable is one of the storing representations and establishing
relations. This mechanism relies on several cognitive heuristics, one of the most rel-
evant being causality. Mental representations connected within each other constitute
maps of cognition. According to the constructivist approach, cognition can be de-
picted as an input-output process with a feedback mechanism of proof and update.
Because of this feedback, which aims to validate the output of the cognitive process,
cognition is involved in a circular confrontation with the external environment and
therefore is articulated in a self-referring way. This acts in the sense as a reinforce-
ment cycle of the kind of “learning by success,”87 according to which “[i]f a pattern
of cause and effect is applied successfully, the perceived accomplishment will fortify
the relations”88 . Success is meant here in a constructivist interpretation, a provi-
sional validity or viability. In particular, this does not require an internal (mental)
representation to reproduce reality exactly, since such a metaphysical ideal would
have no operational content. Instead, a rule simply has to assure that a specific model
is provisionally “valid” or “viable” for the action of an individual who acts in given
conditions.
It is to this extent that the constructivist approach introduces the notion of “cog-
nitive equilibrium” as a sort of measure to which the viability of a mental represen-
tation can be related. An individual can be said to be in “cognitive equilibrium” if
the actions generated by her internal environment are consistent with her objectives,
given the environmental responses.89
To make the concept of provisional viability more operative, the notion of cogni-
tive equilibrium could be related to that of satisficing, since it allows for heteroge-
neous mind constructions to coexist and to meet subjective aspirations. Viability is
based on individual experience. Experience should not be interpreted here as obser-
vation, but as action, since it is only through action that the individual is able to test
her own mental models. The test result operates on the construction of knowledge
as feedback, which leads either to the validation of actual mental models or to their
modification. In this way, knowledge is a representation of its experience rather than
a representation of the world.90 This feedback mechanism enables the representa-
tion to characterise the cognitive process as self-referential because every construct,
85 The perspective of constructivism will be deepened later on Sect. 1.1 of Chap. 3.

86 It will be in particular referred to the work of Neisser (1976).
87 Handlbauer (1997, p. 763).
88 Idem.
89 Idem.
90 Cf. Maturana and Varela (1987).
1.9 Self-Reference in Social and Individual Decision-Making 27
once confronted with the subjective experience of the external environment, shall be
reflected to the mind which originated it, i.e. it shall be self-reflected.
Thus, human cognition is steadily recursively engaged in the elaboration of men-
tal models out of external stimuli and in their evaluation according to the individual
experience of the external world, either to consolidate a (subjective) viable construct
or to modify a non-viable one.
Both cognitive maps and cognition work in a self-referring way. The self-
referring structure of decision making, which can be characterised as a continuous
path-dependent self-referential process (as in Fig. 1.4), adds to the self-referring
structure of cognition.91 Path dependency is both the result of experience, to which
cognitive maps are related for the sake of their validation, and of prior perceptions
and constructs, by means of which provisional validity is weighted. In this sense,
cognitive maps accumulate experiences and prior perceptions in a meaningful struc-
ture.92 Therefore decision-making can be seen as the application of cognitive maps
which have in most cases a random structure.
The meaning of cognitive maps stems from the circularity of learning-by-success
on which they are based, that is in the continuous proof at their adaptability to ex-
ternal environment. Updating and revising of cognitive maps, as well as decisions,
initially focuses on viability, then sense-making. In this regard, it should be noted
perceived results
cognitive map
decision
consequences
environment
Fig. 1.4 The self-referring

structure of decision-making
(Handlbauer, 1997)
91 Cf. Handlbauer (1997, p. 764).

92 From Handlbauer (1997, p. 764).
that plausibility matters much more than accuracy.93 Among other things, this can
be related to basic findings on cognitive dissonance,94 for example, individuals are
prone to seek information that will reduce dissonance and avoid information that
will increase dissonance.95
Heterogeneity of behaviour can easily be observed in real life situations and re-
flects heterogeneity of mental maps and constructs. Such heterogeneity stems from
partiality of human cognition,96 since a purpose oriented individual,97 who relies
on given, limited cognitive heuristics and capabilities, cannot but simply strives for
what she perceives to be necessary to manage her usual environment. So, contin-
gency as well as past experiences determine path-dependency and possibility of
cognitive structures98 and consequently of behavioural patterns.
In order to find a viable course, individuals involved in interactive situations have
to make assumptions on the others and on their behaviour. This implies, in terms
of cognitive maps, that individuals have to add assumptions and beliefs to their
cognitive patterns about the others, about their cognitive maps and their behaviour.
Such assumptions rely both on the interpretation of behaviours observed and on
introspection, since individuals are not able to think the way others do. Hereby,
considering the social position and the role of the interacting individuals may matter,
the structure of sense-making is also based on social context.99
Direct communication is in principle never possible because separated cognitive
entities need a means of establishing communication among themselves. Mediate
communication is a construct which suffers from the limits of subjective under-
standing. This is because the communicated message has to be filtered by subjective
cognition and integrated in one’s own cognitive maps. Therefore, the observation of
the others’ behaviour and introspection are the key mechanisms for inserting beliefs
about the others in an individual’s own map of thoughts.
Even for collective decision-making, continuous proof of viability of the cog-
nitive maps regarding both an individual’s own and others’ cognitive structures.
When such constructs are perceived to be viable, a cycle of mutual reinforcement
may arise.100 The question if such empowerment leads collective maps and shared
cognition to develop remains unanswered. Even if the role of intensive exchange of
information in promoting general acceptance of cognitive constructs can be empiri-
cally confirmed, cognitive structures remain fundamentally affected by individually
specific autopoietic cognitive processing.
In observation of social interaction it has emerged that there are collective ac-
cepted behavioural patterns which rely on coherency between individual mental
93 Cf. Weick (1995).
94 See e.g. Festinger (1957).
95 Cf. e.g. Frey (1986).
96 Cf. e.g. Tamborini (1997).
97 Constructivism interprets individuals as purpose oriented. For more see Sect. 3.1.2 of Chap. 3.
98 The possibility of the cognitive structure indicates the possible coexistence of heterogeneous
cognitive maps.
99 Cf. Luhmann (1984, p. 580).
100 Cf. Handlbauer (1997, p. 766).
1.9 Self-Reference in Social and Individual Decision-Making 29
Fig. 1.5 Simple structure of

individual decision-making in perceived results A
a social situation (Handlbauer,
1997)
cognitive map
A
decision A
consequences A
environment
decision B
consequences B
cognitive map
B
perceived results B
maps regarding the individual herself and those regarding the others (Fig. 1.5). Com-
mon acceptance of mental patterns may also work in a self-enforcing way, as some-
times acceptance can make constructs viable independently of their intrinsic value.
Individuals belonging to a social group or community may benefit in several
ways from the shared acceptance of mental patterns. The benefits include an increase
in the predictability of a social situation, a simplification of individual decision-
making in the specific social context and a reduction in the intensity of cognitive
effort necessary, which in turn makes partial knowledge suffice. Furthermore, social
affiliation promotes the convergence of behavioural patterns by means of elimina-
tion of disrupting behaviours, and works in this sense in favour of the survival of the
social group’s identity.
Those considerations help in understanding the primary role of institutions in
ruling social interactions. According to Scott, institutions consist “of cognitive, nor-
mative and regulative structures and activities that provide stability and meaning to
social behaviour”.101 Institutions are mainly concerned with cognitive stabilisation
among social groups. They create a loop in the individual decision-making which
evaluates the viability of the cognitive maps in the institutionalised environment
(cf. Fig. 1.6). Institutions shape the social environment and get perceived as a fixed
part of it. They do not directly belong to individual cognitive maps but influence
their validation because they work towards cognitive stabilisation by consolidation
of socially viable constructs.
101 Cf. Scott (1995, p. 33).
Fig. 1.6 Decision-making in individually

institutionalised environment perceived results
(Handlbauer, 1997)
individual
cognitive map
individual
decision
individual
consequences
institutions
public reflection
Institutions reduce uncertainty and enable individuals to disregard several alter-

natives that could be potentially possible. Thus, individuals can concentrate on insti-
tutionally accepted cognitive patterns that become routinized through this process.
Institutions are at the same time a product of the social process and a constitutive
part of the environment. They are typically perceived as fixed and durable. The ob-
jectivization of institutions acts in self-reinforcing way, inducing the routinization
of cognitive processes and further encouraging behavioural standardization. Self-
empowerment may however disrupt whenever viability of institutional constructs
and responses get seriously questioned: this induces institutional change, which “is
supposed to occur if the individual cycle of perception and the public cycle of re-
flection can no longer be meaningfully connected”.102
102 Cf. Handlbauer (1997, p. 769).

Chapter 2
Reflexivity of Social Reality
The previous chapter has underlined the polymorphism of self-reference and its
involvement in any form of human understanding, activity and conceptualizing.
As neither social reality, nor its observation or description can be abstracted from
their self-referential character, reflexivity and its implications are of central concern
for social research in general. Accordingly, this chapter focuses on the reflexivity of
social reality and phenomena and discusses some of its implications.
In particular, two different orders of reflexivity can affect social reality: a first or-
der of reflexivity involves social reality per se and consists of the social phenomena
that are self-referential in that they may affect themselves, as they can for example
imply, control, or modify their own dynamic or development. A second order of
reflexivity concerns the “discourses” on social reality, such as social sciences and
theorizing. The present chapter offers an overview on common reflexive social phe-
nomena, while the Chap. 3 focuses on the second order of reflexivity concerning
social sciences and theories.
It can be essentially premised, that the first order of reflexivity, which invests
social reality, depends on the autopoietic character of social phenomena that create
themselves on the basis of their inescapable systemic character. The second order
of reflexivity can be fully appreciated if related to the constructivist observer-
observation scheme, as it will be discussed in more detail in the Chap. 3.
One of the first difficulties that the analysis of the reflexivity of social reality
posits is represented by the huge range of phenomena which can be subsumed un-
der the label “social reality.” Therefore, the analysis will begin with specifying the
notion of “social reality,” which will be conceptualized in opposition to “natural
reality.” After that, some notes on the reflexivity of social reality will introduce
an overview on some common and widely analysed reflexive social phenomena.
This overview does not have the claim of being comprehensive and ranges from an-
thropology, linguistics, law, politics, sociology, and psychology, and then concludes
with examples of reflexivity involving the economic reality, as illustrated by the
dynamic of financial markets and of the business cycle.
This overview is conceived as a natural first step for the successive enlargement
(in the Chap. 3) to the reflexivity affecting social research and theorizing.
32 2 Reflexivity of Social Reality
The difference between reflexive social phenomena and social discourses that
lead to reflexive social phenomena, which has been condensed in the notions of
first- and second-order reflexivity, is not so clear and sharp as one could think at a
first glance. Social reality is in the end by human activity, which is coined by human
thought and consciousness, so that the separation between action and cognition, be-
tween practical and intellectual activity, is something fluent, a continuum rather than
a dichotomy. Reflexivity of social reality is due to the fact that the intentional action
of the individuals defines the course of the social system the individuals belong to.
2.1 What is Social Reality?
Social reality encompasses the human aspects of the world and it is constituted
by tenets, beliefs, principles and opinions which may inspire the behaviour of a
community. Social reality can be better defined negatively, i.e. by distinguishing it
from everything that does not belong to its realm, though from the natural reality.
In the tradition of realism, objective facts can be divided in two categories: nat-
ural and social facts. The first exists and follows its course, independent of human
perception, thus independent of what individuals may think about it, while the latter
due to its relation to human thought, essentially depends on its perception and con-
ceptualization.1 Natural facts, which Searle (1995) expressively calls “brute facts,”
remain the same whether individuals realise them or not, such as whether mankind
exists or not. They have a physical presence that, at least for the supporters of real-
ism, cannot be doubted. The discussion of the eventual ontological priors of natural
facts and if the notion of “objective facts” is tenable at all has been and still is vividly
debated. However, such a discussion goes beyond the purposes of the present analy-
sis, so that it will simply be omitted. A naı̈ve interpretation of the concept “natural
facts” will be adopted, as illustrated by the following example: A piece of paper
has a physical presence, so that it can be depicted as a “brute,” natural fact. How-
ever, independent of the meaning it might assume for a certain community, it may
become a social fact as well. For example, it can be a money bill, if it represents a
monetary value, but it can also be a property certificate, thus representing a property
right. This points to the relation between social and natural facts: social and physi-
cal reality are not completely separated and independent from each other, but they
overlap each other.2 It can be argued that social facts can exist without natural facts.
Social reality is embedded in physical reality although the borders among them are
not always univocal. The questions of whether there can be social without natural
facts and where the demarcation appears between the “brute” essence of reality and
the perception by means of which it can be connoted, are raised. Critical examples
in this sense are verbal agreements, ideas, copyright, etc.
1 Cf. Searle (1995).

2 Cf. Searle (1995, p. 35).
2.2 Recursivity of the Social Reality 33
This short discussion can be summarized when social facts are considered a hu-
man product that are essentially in existence because of human construction. A so-
cial fact gets created when a community of individuals determines a function for a
physical object or for certain circumstances, a community defined as that group of
people who accept the assigned function and conform to it.3
2.2 Recursivity of the Social Reality
Recursive relations play an important role for social phenomena and are almost
pervasive both of social reality and of the social sciences.
Reflexivity of social reality is due to the fact that individuals involved in a social
system act intentionally, for example, they try to reach a certain end state from a
given initial state, and, in the process, define the course of it. So, the individual
mental representations and expectations play a decisive role in shaping the social
reality. Examples of recursive phenomena can affect the different aspects of social
reality and have always attracted the attention of social analysts.
A central notion for appreciating the essential reflexive character of social reality
and recursive phenomena is that of “autopoiesis,” the process whereby a system or
organism produces itself, its own components and by means of which can distin-
guish itself from its environment. Autoiesis indicates in other words self-creation,
i.e. a self-creative act or process and can be applied to the description of a social
system. Social systems are essentially autopoietic, as they define their own identity
and steadily reproduce it in their interaction with the external environment, in order
to maintain the system’s survival. The notion of “autopoiesis” was first introduced
by Maturana and Varela (1973) as an attribute to describe the nature of living sys-
tems, e.g. the biological cell, that are able to produce all the components needed
for the maintenenance of the living system out of an external flow of ressources.4
The application of the concept of “autopoiesis” to social reality can be then traced
back to Luhmann’s work on systemic theory,5 according to which social systems are
autopoietically closed, because they rely on a flow of resources from their environ-
ment in order to continue maintaining their specific identity and to differentiate from
the environment. The self-referring nature of any social system can be related to its
autopoiesis, as the survival and maintenance of social facts relies on the infinitum
regressum of filtering and processing information from the environment.
3 Cf. Searle (1995).

4 An autopoietic machine is defined as “a machine organized (defined as a unity) as a network of
processes of production (transformation and destruction) of components which: (i) through their
interactions and transformations continuously regenerate and realize the network of processes
(relations) that produced them; and (ii) constitute it (the machine) as a concrete unity in space in
which they (the components) exist by specifying the topological domain of its realization as such a
network.” Cf. Maturana and Varela (1980, p. 78).
5 Cf. e.g. Luhmann (1984).
Some elements which corroborate this thesis will be provided by the overview of
reflexive social phenomena which follows.
2.2.1 Reflexivity in Anthropology
The thesis which sustains that language coins and anchors thought is known as
Sapir-Whorf hypothesis, from the name of the researchers who provided it with ro-
bust scientific fundaments.6 In its core argument the Sapir-Whorf hypothesis refuses
the view that language merely mirrors culture and habits and argues that the rela-
tionship between language and thought is one of mutual influence. In other words,
the characteristics and grammatical structures of a certain language influence and
shape the understanding and behaving of its speaking community. This implies in
particular that the mother tongue of an individual has a decisive impact on her way
of thinking and processing information.
Speculations on the reflexive relation between language and thought can be
traced back to vivid debates among Indian linguists up to the sixth century AD.
In Europe, one of the first research studies that contributes to the reciprocity of rela-
tion between language and thought can be ascribed to Humboldt.7 Boas, commonly
recognised as the founder of anthropology in the United States, studied some of
the languages of Native Americans and found that they often belong to different
linguistic families. Sapir carried on Boas’ researches and observed how different
languages could give life to different habits and behaviour. Stating that human be-
ings “are very much at the mercy of the particular language which has become
the medium of expression for their society”8 Sapir stressed the importance of social
constructs, in general, and of linguistic structures, in particular, in shaping the re-
ality in which humans act and interact. Language conventions and habits filter and
orient the interpretation of external reality and therefore influence individual per-
ception and experience of it. This goes so far, that “the ‘real world’ is to a large
extent unconsciously built upon the language habits of a group. No two languages
are ever sufficiently similar to be considered as representing the same social reality.
The worlds in which different societies live are distinct worlds, not merely the same
world with different labels attached.”9
Whorf structured those ideas further, arguing that: “We dissect nature along lines
laid down by our native languages. The categories and types that we isolate from
the world of phenomena we do not find there because they stare every observer
in the face; on the contrary, the world is presented in a kaleidoscopic flux of im-
pressions which has to be organized by our minds - and this means largely by the
linguistic systems in our minds. We cut nature up, organize it into concepts, and
6 Cf. Sapir and Mandelbaum (1986) and Whorf and Carroll (1964).
7 It is here referred to von Humbolt’s essay “Über das vergleichende Sprachstudium in Beziehung
auf die verschiedenen Epochen der Sprachentwicklung.” (Von Humboldt, 1945).
8 Cf. Sapir (1929, p. 69).
9 Cf. Sapir (1929, p. 69).
ascribe significances as we do, largely because we are parties to an agreement to

organize it in this way - an agreement that holds throughout our speech community
and is codified in the patterns of our language . . . all observers are not led by the
same physical evidence to the same picture of the universe, unless their linguistic
backgrounds are similar, or can in some way be calibrated.”10
Whorf’s principle of linguistic relativity criticizes the interpretation of thought
as unilaterally influencing language, that reduces language to a means of expression
of what is already coherently formulated as thought. In spite of that, Whorf’s analy-
sis points at the repercussions that the different grammatical structures of different
languages may have on the mental constructs of a speaking community.
For example, one of Whorf’s most famous research11 focused on the differences
between Standard Average European’s and Hopi’s linguistic structures. It emerged
that while the first linguistic structures tend to analyse space and time in a spatial
static sense, the latter have a more dynamic conception than that, which is based
on processes rather than on points. Whorf argues that this could be responsible for
different understanding of mathematics, of spatial metaphors etc.
Among recent studies dealing with linguistic differences among populations,
which provide evidence for the linguistic relativity hypothesis, Gordon’s 2004s re-
search can be mentioned. This research was conduced on a Brazilian tribe, whose
language only contemplates three counting words, namely one, two and many. The
complete inability of the tribe’s members to learn how to count has been ascribed to
this peculiarity of the tribe’s language and interpreted as a sign of the shaping role
the social construct language may have on the community which originated it.
The Sapir-Whorf hypothesis implies in particular that although language as a so-
cial fact is a construction of the human thinking activity, it may have an influence on
human thought. This indicates that a circular relation between language and thought
establishes and hints at the recursive effects between the two entities. Language
and thought can be therefore interpreted as staying in a self-supporting stabilizing
reference relation, linguistic determinism being its product.
Reflexivity among language and thought has often tickled human fantasy, as
shown by the conspicuous fictional presence of linguistic determinism. In Orwell’s
“1984,” for example, language is interpreted as a means for pursuing political total-
itarian aims. In Orwell’s novel, the idea underlying the formulation of “Newspeak”
is that by abolishing words such as “freedom,” people will no longer strive for it, or
by not knowing the meaning of “revolution” they will never rebel.
As the existence of a social fact depends on the assignment of a certain function
and meaning to certain physical objects or situations (a natural fact) as well as on its
acceptance among a community social group, relations of mutual maintenance and
influence can often occur in the realm of social reality, as the proceed of the analysis
illustrates.
10 Cf. Whorf and Carroll (1964, pp. 212–214).

11 This research was mostly conduced during the 1930s, cf. Whorf and Carroll (1964).
2.2.2 Reflexivity in Linguistics
Reflexivity of the natural language constitutes an interesting case of study and is

worth careful analysis, since it can concern two different levels of the semantic al-
lowance of self-references and of the recursive dynamics of linguistic changes and
maintenance. The first level has been already addressed in 1.5, which focused on the
terms and the constructions through which natural languages admit self-referential
devices and either tolerate the semantic they imply or not. The second level con-
cerns autopoietic dynamics of language and thus addresses self-enforcing and self-
defeating developments that determine the evolution of language. Languages are
constructions of the human thought, but at the same time they can anchor or af-
fect their own “creator,” though they can recursively act on the human thought. The
evolution of a language can be characterised as a circular bi-directional process,
because language evolves as a result of modifications in its speakers’ community
but at the same time may induce some of the modifications that determine its evo-
lution. The conception underlying the considerations on linguistic reflexivity that
follows is that the evolution of a language comprehends both reflexive stabilization
and subversion of norms and can be therefore characterised as the result of those
two opposite circular processes.12
The establishment of a language as the means of communication of a certain
community relies on self-supporting dynamics. This is because the acceptance of
the semantics and of the norms of a language is the necessary prerequisite for a
rough combination of sounds to become a word. Since the living norms of a linguis-
tic community are defined by the convergence of social practices, language norms
represent at the same time cause and effect of this convergence process, which can
then be said to be self-referential.
Conventional usage of language is based on the self-enforcing compliance of the
members of the speaking community. This is because the acceptance of a norm by
the majority or by a conspicuous part of individuals belonging to the linguistic group
of reference represents per se a reason not to deviate but to conform to that norm.
Deviations tend to be sanctioned through misunderstanding and/or criticism, while
compliance is rewarded by efficacy of communication, understanding, social accep-
tance and feeling of belonging to a certain speaking community. These mechanisms
put the basis for the self-stabilization of language, which is a process that ensures
the efficacy of language as a major communication means among individuals. Fur-
thermore, this self-stabilization process stresses how once a linguistic norm is estab-
lished, its survival is caused by the existence of the norm, whereas the existence of
the norm decisively depends on its survival. In this spirit, the maintenance of linguis-
tic norms can be at the same time alleged to the conventional usage of language and
is motivated by such a conventional usage. Linguistic norms can be thus interpreted
as being the creator and at the same time as being created by linguistic conventions
and can be furthermore characterised as “historical, or immanent, or constitutive a
12 Cf. Suber (1989).

posteriori.”13 In other words, the validity of certain linguistic norms refer to cer-
tain historical, immanent conditions and are neither eternally valid nor immune to
changes and they constitute almost paradoxically a posteriori, because they “struc-
ture not only what we approve but what we understand and how we act.”14 Their
“normative priority to experience arises (‘congeals’) from the flux of history and
passes away again.”15
The other dynamics which are involved in the evolution of a language involve
the reflexive substitution of language norms and grants in this way for a reflexive
dynamics of change.
There are many reasons for linguistic changes to take place and correspond-
ingly many mechanisms through which linguistic changes may occur. Lending
words from other languages (both foreign and technical), introducing new words
to cope with new needs and circumstances, “playfulness, imitation, laziness, igno-
rance [. . .]”16 are just few examples of the numerous mechanisms through which
linguistic changes may happen.
In general, the processes by which linguistic norms get substituted, revised or
commuted can be said to be based on a logics of “amendment through violation”.17
Furthermore, the mechanisms through which new language norms are estab-
lished, or through which old norms get changed, can be divided into phonetic and
non-phonetic ones. While linguistic changes happening on a phonetic base can be
lead back to “mispronunciation,” the class of non-phonetic mechanisms of linguistic
change is more articulated and illustrates clearly the role of errors and violation in
the process of linguistic change.
In all languages it is possible to find cases in which systematical mispronuncia-
tion of words induces their spelling modification, so that a word substitution occurs.
A mispronunciation becomes only then systematic if it does not get sanctioned,
whereas sanctioning linguistic deviations has been discussed as an essential base
for a language to gain the status of means of communication. It should be hereby
premised that the notion of “mispronunciation” implicitly requires a standard of
correctness. For a language such a standard consists in a family of acceptable pro-
nunciations. In addition, the authoritative source for specifying the notion of accept-
ability among a certain speaking community is represented by its members, whereas
no hierarchy among them can be in principle justified, if the conventional basis of
language is accepted. This allows for the possibility of a norms violation, whenever
the boundaries of mutual understanding do not get overwhelmed. Mutual under-
standing is a relative notion that can only be specified with reference to a particular
speaking community, whereas speaking communities “overlap each other, admit of
innumerable borderline members, and may take very different shapes depending on
13 Cf. Suber (1989).

14 Idem.
15 Idem.
16 Idem.
17 Idem.
what case of usage we are investigating.”18 Official languages coexist with slang,
dialects, technical languages, familiar lexicons etc.
Examples of language changes due to mispronunciation are the English words
with Old French origin “naperon,” which changed into “apron,” or the consonant
inversion between “s” and “p,” which occurred in “waeps,” which was the original
word for the insect named “wasp.” In several other cases letters have been added
in order to simplify the pronunciation, for example “thunder” did not originally
contain the “d,” showing more clearly its relationship with its German equivalent
“Donner.”19
Non-phonetic changes can stem from popular but wrong etymologies, mislead-
ing backformation, according to which the prior between a verbal and a substan-
tive form get exchanged and consequently some syllables get improperly added or
omitted, or per analogy with other similar words.20 The latter mechanism based on
heuristic reasoning and analogical thinking and is nicely illustrated by the children’s
way of speaking. Children typically infer regularities in the language and they of-
ten apply them improperly, not being able to mind for exceptions and irregularities.
Whenever a conspicuous number of members of a speaking community conform to
the improper application of a rule such a violation may become more widely ac-
cepted, till it might amend the norm. For example, the past form of the English verb
“to snow,” which was originally irregularly built as “snew,” got changed into the
‘regular’ “snowed,” while the original regular past form of the “to strive” (“strived”)
progressively changed into “strove,” which was most likely derived per analogy with
other similar verbs.21
Finally, even the reflexive subversion process of a linguistic norm requires a re-
flexive stabilization process in order to establish ex proprio vigore, so that linguistic
changes can be interpreted as emerging from the combination of loops working in
opposite directions and respectively relying on a self-disrupting and self-enforcing
dynamic.
It can be therefore concluded, that “the mutability of language and its norms is a
result of the balance of two reflexive processes. One is the self-reinforcing stability
we called self-stabilization, and the other is the reciprocal causation and reflexive
hierarchy we see in any norm of vulnerable to change from the posterior usage it
structures.”22 Those opposite processes balance, in the sense of leading to a sta-
ble development, essentially because of their different time horizons: while norms’
constraint is a day to day process, norms’ change typically takes years to occur,
or it takes a period over which changes in the speaking community may become
relevant.
18 Cf. Suber (1989).

19 The examples refer to Suber (1989), to which it can be forwarded for further details.
20 Cf. Suber (1989).
21 The examples refer to Suber (1989).
22 Cf. Suber (1989).
2.2.3 Reflexivity in Law
Self-reference casts its shadows also in law and may work both stabilizing and dis-
rupting factors into the development of a legal system. Paradoxes or infinite re-
gresses might arise, when laws contain self-referential statements, which, as it will
be discussed in this paragraph, cannot be totally avoided. Paradoxes and recursive
dynamics would always have tremendous destabilizing effects if juridical praxis
had not found in many cases pragmatic solutions and routine procedures for es-
caping the contradictions they in principle always imply. Still some self-referring
instances or sentences can actually give life to logical inconsistency that puzzle not
only logicians and philosophers but create shortcuts in the juridical system. Some
other self-references might however even represent a useful way to avoid infinite
regresses.
For example, the principle of validation of each law through its legitimation by
prior or higher law should ensure the legitimacy of a legal system but implies an
infinite regress.23 To escape the shortcuts such an infinite regress would imply, ju-
ridical systems have to admit some laws to be valid ex proprio vigore. A process of
validation by its own strength, by its own validation clearly implies a recursive dy-
namics. This common legal device illustrates a case in which a self-referring clause
acts to correct another self-referring device.
Similarly, according to the so-called “bootstrap doctrine,” courts must have some
forms of self-applicable jurisdiction at their disposal, in order to be legitimate to
rule some questions concerning their own jurisdiction without forwarding them to
another court, which if put in the same situation would have to forward them to
another one and so on.24
Self-legitimating laws and courts ruling on their own activity represent cases in
which recursive circular structures are employed to interrupt different potentially
destabilizing loops and to grant the smooth functioning of a juridical system.
As the rule, however, legal circles are more likely to create problems of un-
groundedness, instead of solving them. A “renvoi” occurs if a court has to consider
and to deliberate on the law of another state.25 Of course, even if in principle a ren-
voi or every other circular remand cause a loop that cannot be escaped or solved,
juridical systems are usually able to find a pragmatic solution to avoid jurisdictional
stagnation. Perfect legal norms and perfect contracts cannot exist, since it is not pos-
sible or at least not economically sustainable to foresee all possible future states and
developments. Therefore, legal systems are always provided with residual clauses
that enable them to solve conflicts in a finite time and with a reasonable disposal of
resources.
23 Cf. Suber (1990).
24 Cf. Suber (1990).
25 This could happen for example when a contract that has been stipulated in a certain state gets
violated by one of the parts involved in another state. There are cases in which, despite the contract
explicitly states that such controversies are competence of the state in which the violation takes
place, the laws of that state want such controversies to be solved according to the laws of the
countries in which the contract has been stipulated. For further examples see Suber (1990).
Another formally controversial matter in law which is based on a reflexive mech-

anism is the constitutional amendment. The concept of “amendment” is used to
depict the alteration of a law, whereas legal systems with “rigid” or “entrenched”
constitutions rule the amentment of their constitutional law through a specific pro-
cedure, which differs from the one adopted for the other laws.
An amendment needs an amending clause to be formally admitted by a system.
Because of its universality, any amending clause is feasible due to its recursive ap-
plication. Therefore, self-amendment is in principle always possible. This resembles
the so-called “paradox of omnipotence:”26 a deity to be as such has to be omnipo-
tent but can she create a so heavy stone she cannot lift? It can be similarly argued to
what extent an amending clause can amend itself.
Self-amendment of constitutional norms stems from the need to balance between
the two opposite instances of the stability of a legal system as a product of its formal
consistency and the possibility of its modification., on the one hand to overwhelm
shortages that may follow from the incompleteness of law and on the other hand to
cope with new necessities.
2.2.4 Reflexivity in Politics
The logic underlying amending clauses in laws reflects the necessity of limiting
power in modern political systems. Political power derives from sovereignty and
is exercised either directly or through delegating representatives by the sovereignty
holders. Every modern political system has to rule both the self-limitation and the
self-augmentation of the political power. This involves the definition of limitations
the political power has to underline, both concerning the time (e.g. time for the
representatives to govern) and the content of its exertion (e.g. inviolable rights).
The political course may be affected both by stabilizing and disrupting recur-
sive dynamics. For example, a shared fundamental ideological background under-
lies each political system and creates a self-validating system of beliefs that works
self-reinforcing and self-isolating.27 This shared ideological background - exempli-
fied by national symbols, flag and patriotic creed - increases cohesion and sense of
belonging among the members of the political system and promotes the acceptance
and maintenance of its institutions. Such an ideology is an expression of a certain
political system and affects at the same time the system’s dynamics in a stabilizing
way. Similar considerations obviously apply to every kind of ideology, so that ide-
ologies can be interpreted as self-enforcing systems of beliefs.28 Ideologies can also
work in a destabilizing way on a political system, leading to structural changes that
might even assume a revolutionary connotation. This indicates that revolutions can
26 Cf. Suber (1990).

27 Cf. Bartlett (1989, p. 15).
28 Similarly, also religious belief-systems can be said to be self-referential. Cf. Bartlett (1989,
p. 15).
also be interpreted as reflexive phenomena, in concomitance with whom the internal

dynamics of a political system becomes self-destructive.29
The legitimacy of a political system and of its institutions is indissolubly tied to
its perception among the members of that political system. The system’s stability as
well as the stability of its institutions may also depend on the coherence between
reality and its perception, so that a significant gap between reality and perception
might erode legitimacy of institutional arrangements. This implies that in certain
situations, even without objective relevant social changes, the perception of certain
institutional arrangements as inadequate might, in itself, spread among the members
of the political system and call for their modification.
From the discussion of some of the reflexive phenomena affecting politics it
emerges again how self-referential phenomena involving positive feedback loops
work in a stabilizing self-reinforcing way, allows for continuity and does not intro-
duce contradictory elements, whilst self-referential phenomena which rely on nega-
tive feedback loops have a destabilizing character and challenge logical thinking.
The balancing between positive and negative self-referring dynamics represents
a particularly critical issue in the perspective of a political system. Similarly to the
processes of linguistic change, subversive political dynamics owes its survival to a
successive stabilization process, whose regulative efficacy depends on the one hand
on the degree to which it self-enforces and on the other hand to the disruptiveness
of the initial negative loop. Even if gradual changes can also lead to deep modi-
fications in a political system, more radical changes can be needed in some cases,
especially when it comes to subverting the balance of power or the established peck-
ing order, as illustrated by the famous quote “If you strike at a king, you better
kill him.”30
2.2.5 Reflexivity in Sociology
Within the social sciences, sociology is maybe the only one that directly attempts to
confront the problems of reflexive social phenomena. This is because sociologists
are conscious that social phenomena might occur simply because of its common
acceptance among a certain community. The existence of social facts is tied with
their perception among the social actors and socially relevant developments which
might decisively depend on the degree with which they are thought to be true and
accepted.
The difference between the two orders of reflexivity that can affect social reality
becomes particularly slight if applied to the realm of sociological analysis. Sociol-
ogy deals with all social phenomena, with discourses about social reality included.
Sociology focuses both on social facts and on the opinions and beliefs the social
29Cf. Bartlett (1989, p. 15).

30Cf. Holmes (1980), who ascribes this sentence to the philosopher, essayst and poet Ralph Waldo
Emerson.
actors hold about them. The constitutive role of opinions and beliefs in shaping so-
cial reality represents a core topic in sociology. For example, the research program
of the sociology of scientific knowledge aims at the analysis of social influences
on science and addresses explicitly reflexivity in sociology. To simplify things, one
main thesis underlying sociology of scientific knowledge is that social factors play
an active role in shaping the development of science.
Plenty of examples of self-fulfilling as well as of self-destroying dynamics can be
mentioned. A published prediction may affect the predicted event and either work
toward the self-fulfilment or the self-destruction of the prediction. The disclosure
of a public opinion survey or the publication of scientific results can purposefully
strive for certain reactions among the public.
While reflexive predictions will be more specifically approached in the next chap-
ter (par. 3.2.3), the reflexive effects of beliefs and opinions will be here illustrated
by the so-called “Pygmalion effect.”
The Pygmalion or Rosenthal effect refers to situations in which pupils, who are
expected to perform better than others, will indeed perform better. This effect was
first examined in a study by Rosenthal and Jacobson (1968, 1992), in which some
teachers were misleadingly told that some children had a higher-than-average IQ. It
was shown that the expectations of the teachers led the children to an actual en-
hancement of their performance. The name refers to Ovid’s tale of the sculptor
Pygmalion, who created a statue of perfect beauty and fell in love with his own
creature.
As said, the thematic of reflexive predictions will be more specifically ap-
proached in the next chapter (par. 3.2.3).
2.2.6 Reflexivity in Psychology
Psychological states of individuals can have a decisive impact on reality, as different

beliefs or attitudes might ceteris paribus lead to the realisation of opposite develop-
ments and produce alternative outcomes.
For example, the cognitive bias in which a researcher may occur, if she expects
certain results can be mentioned: the researcher could either unconsciously manip-
ulate the research method and the experiment or misinterpret the data and therefore
actually obtain the expected results. This bias is known as the “observer-expectancy
effect.”31
Rosenthal (1998) provided evidence that an experimenter can subtly and unin-
tentionally communicate the experimental subjects’ own expectations and biases
through a conspicuous number of cues or signals. These cues can significantly af-
fect the outcome of the experiment and artificially create the conditions for expected
developments to occur.32
31 This effect is also known in the literature as “experimenter-expectancy effect,” “experimenter
effect,” or “observer effect.”

32 Cf. Rosenthal (1998).
This effect is a clear example of how even science, despite all methodological
premises that should grant for the objectivity of enquiry and results, cannot tran-
scend its own dimension of human enterprise and is therefore “held in check” by the
consciousness of the human mind. In other words, the reflexivity of human under-
standing implies the reflexivity of all its elaborations. Scientific knowledge makes
no exception.
A possibility of debiasing the observer-expectancy effect is to rely on a double-
blind methodology. This consists in concealing both experimenters and subjects
from which subjects are assigned to which group (control and test group) until the
end of the study. This procedure can best be applied in computerised experiments
but also presents the disadvantage of being quite costly.
Also the participants to an experiment or to a research study can manipulate the
results of the analysis. In similar cases it can be spoken of “subject-expectancy ef-
fect.” Plenty of evidence of this effect can be found in psychotherapy and medicine.
The effects of the subjects’ expectancy on the efficacy of a certain cure can even
imply healing processes to be accelerated or to fail. That the patient’s symptoms
can be alleviated, just because of the belief in the efficacy of an otherwise ineffec-
tive treatment, is the well-known placebo effect. The opposite effect, the so called
“nocebo effect,” can occur as well, if a patient disbelieves an effective treatment.
The usual procedure to prevent biased behaviours of the experimental subjects
consists in running single-blind trials, i.e. not to reveal to the subjects whether they
belong to a test or to a control group.
When the experimental results confirm the experimental hypothesis, but not for
the expected reasons it can be spoken of Hawthorne effect. The assumption underly-
ing this effect is that the results simply occur because of the subjects’ awareness of
participating in an experiment, with no other plausible explanation. The Hawthorne
effect is thus a reaction of the experimental subjects to the fact that they are objects
of analysis. It therefore represents a case in which the act of observing induces a
modification in the observed entity that would not have taken place without the ob-
serving act and provides evidence for a recursive causal relation between the act and
the object of an observation to establish.
The Hawthorne effect first emerged from investigations on the influence of dif-
ferent work environment characteristics on productivity, which were conducted be-
tween 1924 and 1932 at the Hawthorne work of the Western Electric Company in
Chicago.33 It could be observed, that the modification of different factors of the
work environment, e.g. pay, light levels, rest breaks etc., all induced an increase in
productivity, independently on the direction of such changes (increase or decrease).
It was also found that even the return to original conditions had a positive effect on
the productivity and therefore any plausible explanations could be formulated except
that the awareness of being an object of a study per se motivated the experimental
subjects to modify their behaviour.
33 These studies were conducted by Fritz J. Roethlisberger and William J. Dickson.

2.2.7 Reflexivity of Economic Reality
Reflexive dynamics can also affect the course of the economy, as the actors inter-
acting in an economic system act purposefully and can on this basis have a part in
determining economic events or developments. It can be thought of several exam-
ples of self-fulfilling as well as of self-destroying dynamics concerning economic
systems and interactions.
For example, the disclosure of a market research can influence both investment
and consumer behaviour. It can assume a different predictive value independent
of its dissemination status and of how trustworthy it is perceived to be among the
economic actors involved.
Similarly, the German Federal High Court recently had to decide on the responsi-
bility of the Deutsche Bank for the bankruptcy of Kirch’s corporation; Kirch accused
the bank of having caused its bankruptcy by publicly doubting its creditworthiness.
It can be said, that the expectations of the market’s participants can have a deci-
sive influence on the market’s performance. Evidence for similar dynamics can be
observed in many contexts. Examples can be found that inflationary and deflationary
developments often show a self-fueling character, and that the compounding of rein-
vestment can influence economic growth. Reflexive dynamics can also be exploited
as a way of achieving economic goals, as illustrated e.g. by self-investment as pos-
sible management strategic device. They have also been contemplated to different
extents from a theoretical point of view, as attested e.g. by game theoretical pre-
scriptions that can reveal both a self-undermining and self-guaranteeing character.
2.2.7.1 Reflexivity of Finance
It is unquestionable, that the development of the financial markets is influenced and

mirrors the historical course of the events: political stability or shortcuts, profound
social changes as migrations, conflicts etc. inevitably reflect on the dynamics of
financial markets and decisively coin their functioning. However, because the mar-
kets are a product of the interaction of purpose oriented actors, they can sometimes
anticipate real developments and therefore enforce or accelerate their occurrence.
Financial interactions are social facts that are embedded in the social reality, which
means that mutual influences between them and political, social as well as cultural
events and developments can never be excluded. This implies that causal relations
between financial interaction and other social facts have to be interpreted as a circu-
lar recursive process.
Soros’ (1994) conception of the financial markets and their functioning focuses
exactly on this point, namely on the relation of mutual influence in which financial
markets stand with the social system they are embedded in. This has been quite fre-
quently discussed in financial literature. It is commonly known and accepted both
among managers and operators and among researchers and analysts34 that manag-
34 Cf. Copeland, Koller, and Murin (2000, p. 52).

ing expectations of markets’ participants can have a decisive influence on market’s

performance.
This is also illustrated by the three forces that are typically considered to govern
the market’s course. These forces are the fundamentals (revenue, cost, capital etc.),
the technicals (price and volume data) and the market behaviour, whose dynamic is
not easy to predict.
Operating in a financial market requires thinking about the system and the events
one is participating in. The thinking contributes to shape the object of thought, which
is social reality “not independently given, but contingent upon our decisions.”35 As
thought constitutes an element, a part of the social reality, it should be considered in
examining social dynamics. Since “our thinking forms part of the reality we think
about, the separation between thinking and reality is breached. Instead of a one-way
correspondence between statements and facts there is a two-way connection.”36 In
this spirit Soros criticizes the view that the market should always be right. In Soros’
eyes the market can only be wrong, as it is always, inevitably biased. The market’s
distorted view of future events works in two directions. First, market participants
rely on such a distorted view. Second, this view can influence its future develop-
ments. The impression that markets predict the truth can actually be inferred. But
only because of this bidirectional influence does the impression appear to be sus-
tained by evidence. This relation of mutual influence implies for the one side, that
the view of the market’s participants contributes to shape the course of the market,
while from the other side, concurrently the course of the market enforces or corrects
the participants’ view.
Soros (1994) applies the concept of “reflexivity” to the description of this mutual
relation of causality between the course of the market and its participants’ expecta-
tions. In Soros’ view the application of the methodology of the natural to the social
sciences is condemned and its failure has to be ascribed to this reflexivity. As eco-
nomics is the social discipline with the highest ambition of resembling formalisms,
axiomatic construction and method of the natural sciences, the effects of reflexivity
might have devastating consequences on its scientific validity. This is particularly
evident if one considers the notions of rationality and perfect understanding, which
are assumed to inspire the behaviour of the economic actors. These assumptions are
clearly inadequate to stylize the real individual cognitive capabilities that can ac-
tually be observed. Imperfect understanding, limited processing and computational
capabilities seem to inform the behaviour of individuals who deal with economic
problems and interactions.
Financial markets are always biased, which is confirmed by the continuous fluc-
tuations that steadily perturb their long term dynamics. The force acting beyond
such perturbations is, in Soros (1994) view, mainly constituted by the participants’
perception or preconception about the dynamics of the market.
The reflexive relation, which links the thought of the actors involved to the situ-
ation they are part of, can be expressed by means of a system of recursive functions
35 Cf. Soros (1994).

36 Cf. Soros (1994).
consisting in a “cognitive” or “passive” function, y = f (x), and a “participative” or

“active” function, x = φ (y).
The cognitive function expresses the participants’ perception and opinion on the
market, for example, y, as a function of the independent variable, x, which summa-
rizes the main features of the market, i.e. the trends which underlie the market. In
this way, the function y points out that the markets’ participants form their beliefs
and opinions on the dynamic of the market on the basis of the observation of the
market. The beliefs as a result of the thinking activity of the individuals will then
reflect on the dynamic of the market. Therefore, participants’ beliefs constitute the
argument of the participative function, x = φ (y), which thus catches their influence
on the course of the market.
Such a recursive system of functions ensures that changes in the cognitive func-
tion are captured by the participative function, which describes the course the events
will take. Let y = f (x) be a “cognitive” or “passive” function, which expresses in-
dividual beliefs about the situation that they are confronted with and x = φ (y) a
“participative” or “active” function, which catches the repercussions of such beliefs
on real developments.
The reflexivity of the system constituted by y = f (x) and x = φ (y) can be appre-
ciated by substituting the arguments of the two functions with each other, so that
both y = f [φ (y)] and x = φ [ f (x)] become a function of themselves. This is illus-
trated by Fig. 2.1, which points out that the market’s underlying trends, x, influence
the participants’ beliefs, y, as expressed by the cognitive function y = f (x), while at
same time the participants’ beliefs, y, affect the market’s trends, x, as captured by
the participative function x = φ (y).
This reflexive functional system can either develop toward the equilibrium or
generate changes and invert trends. While pre-programmed routinized cognitive
patterns and procedures can be expected to inform the decisions of the market’s
participants in usual settings, historical events are most likely to induce significant
changes in the participants’ beliefs. Therefore, the participants will act in a stabiliz-
ing way toward the market equilibrium when they perceive the course of the market
to be normal. They will react and contribute to the deviation from the equilibrium,
Cognitive Function
y= f(x)
Market’s trend
Participants’ beliefs
(x)
(y)
Fig. 2.1 Reflexive interaction

between cognitive and par-
ticipative function (author’s Participative Function
representation) x= j (y)
whenever they perceive the market to be challenged by unusual developments. This

yields for the coexistence of stabilizing and destabilizing reflexive perturbations, the
combination of which inform the course of the social reality.
The stock market represents an ideal setting for analysing reflexivity and appre-
ciating its biasing influence. This is because the stock market resembles the basic
feature of a market with perfect competition: it is a centralized marketplace, with a
high number of buyers who do not make the price and do not have a direct imme-
diate control on it. In addition, in the stock market entrance barriers are practically
absent. Homogeneous goods are traded with very low transaction costs, almost in
absence of transport costs. Furthermore, brokers’ reports provide a plausible basis
for inferring something about the beliefs that are held by the market’s participants.
These reports can convey a picture of largely shared opinions about the market in a
certain moment. However, it is, in principle, impossible to understand exactly which
beliefs and opinions are mostly shared by the market’s participants. It is even more
difficult to infer such beliefs ex post on the basis of the observation of how the stock
market has developed. It can be inferred that positive optimistic beliefs must have
been shared by the majority of the markets’ participants, if ceteris paribus stock
prices grew. On the contrary, the fact that ceteris paribus stock prices failed can be
ascribed to negative pessimistic beliefs and expectations. However, even the speci-
fication of the ceteris paribus condition is only possible on the basis of contrafactive
speculations and therefore, different ways to infer the participants’ beliefs have to
be explored.
The course of the stock market is illustrated by the development of the stock
prices. This is again shaped by an “objective” and a “subjective” component. While
the first component is represented by the development of fundamental variables
of the market, the latter reflects the prevailing beliefs of the market’s participants. In
this perspective, the dynamics of the stock market can be simplified to the product
of interaction between the trends in the fundamentals and the prevailing opinions
and beliefs that are held by the market’s participants. The development of the stock
prices - from boom to collapse – can be related to the interaction between the fun-
damentals trend and the prevailing market participants’ opinion. In this way, the
reflexivity that influences the market will be closer examined.37
Reflexivity can be extrapolated from the comparison between the development of
the market’s and of the fundamentals’ trends. Reflexivity can be interpreted to work
in a stabilizing way whenever those trends reveal similar dynamics. Reflexivity can
be alleged to have biased the course of the market, whenever there is a discrepancy
between the market’s underlying trend and the trends in the fundamental.
Under the label “fundamentals” are summarized all those variables which may
affect the market. Among them are e.g. dividends, patrimonial values, cash flow and
earnings. For simplicity, in the following model of the stock market, the earnings
per share have been chosen as representative for the fundamentals because they can
be interpreted as a variable which results from the market’s development.
37 Cf. Soros (1994).

Stock Prices
Earnings per
Share
Money
A B C D E F G H I
Time
Fig. 2.2 Reflexivity in the Stock Market (Soros, 1994)
Therefore, the stock market’s development will be modelled (see Fig. 2.2) by
observing the development of stock prices (as result of the market’s course) and the
development of earnings per share, which have been chosen as representatives for
the fundamentals.
Typically, market cycle is articulated over the following steps. At the beginning
(AB in Fig. 2.2) a certain trend positively affirms and affects the per share earn-
ings. Most likely, market’s participants hold spread beliefs about the market’s future
developments, so that no prevailing opinion can be outlined. As time passes, the
market’s participants gain awareness on the developing trend and enforce it through
conform expectations (segment BC in Fig. 2.2). The trend further develops with a
certain stability and is not decisively disturbed by slight transitory changes that can
occur in the participants’ expectations, as illustrated e.g. in CD. The stability and
persistence of the market’s trend finally breaks participants’ last resistances and en-
forces trend conform expectations. Expectations remain optimistic and sustain the
stock prices in spite of a deceleration in the earnings’ trend (DE). In EF partici-
pants’ optimistic expectations are the only force sustaining the stock prices and are
no longer confirmed by reality, as it is illustrated by the inversion on the fundamen-
tals’ trend. In this insight, “The essence of a speculative bubble is a sort of feedback,
from price increases, to increased investor enthusiasm, to increased demand, and
hence further price increases. The high demand for the asset is generated by the
public memory of high past returns, and the optimism those high returns generate
for the future. The feedback can amplify positive forces affecting the market, making
the market reach higher levels than it would if it were responding only directly to
these positive forces.”38
Eventually, the market’s participants recognise their optimistic expectations to
be erroneous. This then induces the prevailing opinion to change (FG). The stocks
38 Cf. Schiller (2001, p. 3.)
loose their last support and their price fails, whereas both the expectations and the
fundamentals’ trend work in the same negative direction and amplify each other’s
effect (GH). It can be observed, that “Just as the euphoria of a boom exacerbates
investors’ preference for making abnormally large returns, a bust exacerbated the
despondency of suffering deep portfolio losses.”39 Finally, as the pessimism of the
markets’ participants becomes too big, the market slowly stabilizes (HI).
The cycle that has been sketched is just one of the infinite possible configurations
the dynamic of the stock market can assume. On its basis, the interaction between
beliefs and expectations of the market’s participants and the market’s “objective”
development can be observed. In this frame, reflexivity emerges as the residual force
that affects the market’s dynamics and can find expression in the difference between
the fundamentals’ trend and the market’s course.
In “The Alchemy of Finance,” Soros systematically applies this model of reflex-
ivity to the analysis of several cases, e.g. to the currency market, to the Real Estate
Investment Trusts, to the venture capital boom, and to the conglomerate boom.
The boom that many conglomerate companies experienced in the late 60s was
based on a very simple mechanism, the rising of per share earnings which should
have reflected the good management of a company. But in the end, realty was pro-
duced by the acquisition of companies with typically high dividends but low price-
earnings ratio.
Because of the positive expectations of investors, the shifting of the conglomerate
from high tech to consumer goods does not get reflected in a lower price-earnings
ratio. The positive expectations motivate the investors to buy more stocks and this
further artificially sustains their price.
In more detail, when a high tech company with high price-earnings ratio starts
diversifying, it typically acquires consumer goods companies with high dividends
but low price-earnings ratios. This makes the earnings of the conglomerate company
rise and improves its market value per share. This further increases the borrowing
possibilities of the conglomerate company for further expanding, so that further
consumer goods companies can be acquired. The continuous growth of the per-
share earnings of the expanded company will induce the investors to buy, attracted
by the high price-earnings ratio. The investors’ behaviour will further enforce such
a blown up upward trend, which is only supported by the overall earnings ratio and
does not reflect the fraction of the conglomerate company that operates in low price-
earnings ratio businesses. When eventually the investors realize this, the stock price
will fall to match the real characteristics of the company.
Finally, Soros’ analysis is centred on the self-referential character of the financial
markets. Their dynamics can therefore be described by a reflexivity theory, accord-
ing to which the market process is determined by a two-way feedback loop between
fundamental values, on the one hand, and subjective beliefs and estimations of the
market’s participants, on the other hand. Soros focuses on the interaction between
cognition and action of actors involved, which yields for reflexivity.
39 Cf. Calandro (2004, p. 53).

2.2.7.2 Reflexivity of Business Cycle in Politicised Markets
The dynamic of the business cycle is shaped by the interaction between the objective
features of the economic system, i.e. its fundamentals, and their subjective percep-
tion by the participants in the economic system. This interaction informs market
behaviour. In this perspective, the business cycle can be seen as steered by the ad-
justments between the fundamentals and the way they are subjectively perceived
and processed by the market’s participants.
Par. 2.2.7.1 sketched the dynamic of a typical boom-bust cycle on the stock
market, which has been approximated as being substantially free from “external”
interferences. Its course has been interpreted as mainly determined by its own fun-
damentals and the decisions of the people engaged in the exchange. However, the
inference of politics on economics cannot be understated, as it creates the institu-
tional frame in which economics develops.
Political intervention on economics pursues precise aims and interests and can be
motivated both by regulative purposes and opportunistic instances, e.g. re-election.
The reduction of the interest rate below its market-determined level is a frequent
example of a political device which serves opportunistic concerns. This induces
the market’s participants to spend more which then leads in the short-term to a
boom. The boom is blown up by the increased selling amount, which continues
until the market’s participants finally realize the discrepancy between the biased
fundamentals and those that would reveal the real condition of the market.
The boom-bust model presented in par. 2.2.7.1 will be sketched again and dis-
cussed together with some of the typical political interventions that may occur in
each of its phases. Hereby, the reflexive implications of economic policy making
will be focussed, as they emerge when political intervention creates mechanisms
that affect the political course either in a self-enforcing or in a self-defeating way.40
The market’s participants initially need some time to recognize the trends in
the fundamental. Since the electors typically consider a positive conjuncture to be
favoured by the political action, a good trend in the fundamentals (as illustrated by
segment AB in Fig. 2.2), this typically yields the politicians an increasing popular-
ity. The politicians are as a rule reluctant to waste their potential of interventions in
the very beginning of a boom, where positive further developments of the business
cycle can be expected. In addition, it is in this initial phase that the central bank
typically intervenes by diminishing the interest rate. However, as soon as the end
of the legislature approaches, the politicians’ incentives for adopting business cy-
cle encouraging manoeuvres are high and therefore similar interventions (e.g. tax
reductions) are most likely to be considered even in the initial phase of the busi-
ness cycle.
Segment BC in Fig. 2.2 illustrates price appreciation. It drives market’s partici-
pants’ expectations to grow and the political leadership typically is enjoyed over a
spreading consensus. Despite some perturbations (cf. CD in Fig. 2.2) the positive
trend holds. The upward dynamics of the business cycle continues even when the
40 Cf. the analysis of Calandro (2004).

fundamentals’ trend inverts downward as due to the only support of the market’s
participants’ unrevised positive expectations.
In this phase, politicians and institutions can be interested in artificially sustain-
ing such positive expectations and could for this reason choose to affect some fun-
damental substitutes in order to influence the investment climate.
To do that it can sometimes be enough to encourage certain accounting practices
to spread, as for example focussing on spurious gain indicators in order to induce
an overestimation of the real profits.41 Similar devices achieve the artificial sustain-
ment of the cycle’s upward dynamics, in that they induce the economic actors to
“replace fundamental analysis with fundamental substitute analysis that will sup-
port the boom.”42
In similar settings expansive manoeuvres of economic policy are likely to be
undertaken, e.g. credit expansion. The boom could then be ascribed to wealth effects
created on the basis of the new economic manoeuvres or even to new economic
conditions,43 as for example the debate on the new economy pointed out.
Finally, the market’s participants recognise the prevailing negative course of the
market. Their expectations reconcile with the fundamentals’ trend and as a conse-
quence the price fails uncurbed (see segment FG). Typically and in particular in
those cases where the boom had been artificially amplified, politics should interfere
in order to avoid a market crash.
A contraction in the central bank’s monetary stance, e.g. an increase in the inter-
est rate, then sets a bust in motion. As such a bust develops expectations, it becomes
more negative. The market behaviour gets more and more biased toward what can
be called “irrational despondency”44 and this leads to massive disinvestment and
mass selling (GH). Finally, the market stabilizes progressively, while prices and
fundamentals adjust toward their pre-bubble levels (HI).
41 Cf. Hayek (1970, p. 95).

42 Cf. Calandro (2004, p. 52).
43 Cf. Hazlitt (1996, p. 158).
44 Cf. Calandro (2004, p. 62).
Chapter 3
Reflexivity and Predictability of the Social
Sciences
Reflexivity is inevitably involved in all social facts and discourses about social real-
ity because social discourses are inevitably made by an observer who is part of the
system of observation. Discourses on social reality are elaborations that transcend
meta-logically the reality they focus on, as they are formulated by a subject that
cannot be disentangled from the observed object.
This chapter addresses the unavoidable reflexive nature which characterizes all
kinds of discourses about social facts and reality. This can be ascribed to the in-
volvement of the observer in the system she observes and constitutes an inescapable
condition in which all human conceptualizations concerning social facts are caught.
In set-theoretical terms, human discourse’s self-referring character is given because
its domain (the subject of the analysis) is embedded in the range to which it refers
(the object of the analysis).1
Reflexivity of this kind, which is implied by the so-called “observer-observation
problematic,” constitutes a peculiarity of the social sciences and finds no real
equivalent in the natural sciences. Its most similar problem is represented by the
Heisenberg’s uncertainty principle.2 This can be traced back to the observer effect
and ascribed to the interference of the act of analysing with the analysed object.
Starting from these premises, constructivism extends the effects of reflexivity to any
form of human analysis because the act of human observation inevitably concurs to
shape the object to which it refers. Constructivism states that all that can be experi-
enced or thought is “constructed,” so that reflexivity constitutes an inescapable basis
of all that can be thought and conceptualized. Reflexivity belongs therefore to one
of the central concerns of the constructivist analysis and represents a key concept
for deepening human cognition and behaviour.
Therefore, the first part of this chapter focuses on the perspective of radical
constructivism. Its essential fundaments will be illustrated discussing the observer-
observation scheme and the construct of the self. The specified constructivist per-
spective will then be adopted for discussing and modelling the cognitive processes.
1 Cf. Sect. 3.2 of Chap. 1.

2 This principle will be briefly addressed at p. 68.
54 3 Reflexivity and Predictability of the Social Sciences
Some implications of reflexivity for social research and economics, as they emerge
from the approach of constructivism, will then be addressed.
The second part of the chapter addresses the implications of reflexivity of the
social sciences for the predictability of social and, in particular, economic reality.
After an introduction on different approaches to social predicting, the processes of
explaining and predicting social reality will be compared. The self-altering, reflexive
effect of social predictions which will emerge in this insight will then be developed,
and reflexive predictions will be discussed.
3.1 Constructs and Reality
On a toujours cherché des explications quand c’était des représentations qu’on pou-
vait seulement essayé d’inventer (Valéry)3
The concept of “constructivism” bears a plurality of different approaches
which extend to several disciplines and fields of analysis. One of the first forms
in which constructivism has been formulated is represented4 by the sociological
theory of social constructionism. This theory has been developed based on Hegelian
ideas by Durkheim and then became prominent due to the work of Berger and
Luckmann (1967).
As it is suggested by the word “constructivim,” its different approaches all share
the thesis that all that can be experienced or thought is “constructed” and does not
necessarily reflect any external transcendent real facts.
The introduction to constructivism which follows refers to the radical construc-
tivist tradition, whereas elements of operative5 and of social constructionism6 will
be combined. In opposition to a realist stance, constructivist analysis shifts the at-
tention from reality to its subjective perception and reveals therefore a pronounced
epistemologic rathen than ontologic orientation. The main thesis on which construc-
tivism is based is the unsondability of reality by human knowledge. Reality cannot
be permeated by human knowledge, since it is based on mental representations,
which are constructed out of subjective experience and therefore do not necessarily
refer to ontologic priors. That individuals elaborate mental representations out of
their subjective experience is corroborated by neurological evidence on the activity
of the human brain.7
Radical constructivism fundamentally relies on the following three theorems,
namely the “observer,” the “construction,” and the “validity theorem.”8
The “observer theorem” states that knowledge is essentially a human construc-
tion, so that it does not make sense if disentangled from the human constructive
3 We have always sought explanations when it was only representations that we could seek to
invent (author’s translation).

4 For an introduction on costructivism, see e.g. Fischer (1995).
5 Cf. e.g. Luhmann (1984) and Maturana and Varela (1987).
6 Cf. e.g von Foerster and von Glasersfeld (1999) and von Glasersfeld (1995).
7 Cf. e.g. Hoyenga and Hoyenga (1988) and Haberlandt (1998).
8 Cf. Rusch (1999, p. 8 ff).
3.1 Constructs and Reality 55
activity which generated it. As “there is no knowledge without a knower” and “the
knower personally participates in all acts of understanding”9 knowledge is con-
ceived by the subject who elaborates it. It therefore constitutes an observation that
as such inevitably reflects the subjectivity of its observer. As an observation is in-
fluenced by the subjective way of perceiving and processing the external stimuli, by
past and contingent experiences of the observer, as well as by the boundaries of the
subjective cognition, it cannot be appreciated without considering its observer.
The “construction theorem” posits that knowledge is only individually possible.
This is because knowledge gets constructed by individual cognitive instruments and
repertoire which are specific and highly subjective, so that it will reflect the subjec-
tivity of the knower who carries its limitations and mirrors its experience.
The “validity theorem” underlines the relativity of knowledge and its subjectiv-
ity. Because of the highly subjective character of knowledge, as it is posited by the
first two theorems, the realm of its validity is also limited to the the individuality
of the knower, of the observer. Validation or invalidation of constructs can be inter-
preted only in a subject-related way, so that a construct can only be assessed as valid
or not if related to the individual and to the concrete circumstances from which it
originated.
These theorems enable the profile to have the specific epistemological stance
of radical constructivim. It can be inferred that knowledge is considered to be ar-
ticulated parallel to reality. It is only able to penetrate reality in an “operative”
way and not in an ontologically consistent way. Constructivism thus redefines the
role and meaning of knowledge, which lowers the epistemological capabilities of
the individuals. In a constructivist perspective such capabilities no longer convey a
consistent picture of reality. At most they enable the formulation of “operational”
knowledge, i.e. of knowledge that aims at supporting human interaction with exter-
nal reality. In other words, knowledge primarily reveals an operative nature because
it enables the individuals to deal with the environment they are confronted with by
means of defining cause-effects relationships that build the mental representations
and models.
In a constructivist approach the essence of the facts and reality is relative to the
act of observation on which their definition relies. In other words they do not exist,
if not in the way they are processed by the mind of their observer. While the most
radical constructivist positions refuse to believe in the existence of an objective real-
ity, other currents have a more agnostic stance and merely question the permeability
of reality by the human act of knowing.10 Despite these differences, the centrality
of the act of observation in defining reality brings together the different interpreta-
tions of constructivism. Observing reality is a creative act which implies a “facere,”
an activity which is constitutive of reality and creatively defines factuality.11 Con-
sequently, the concept of truth or at least its accessibility cannot be thought but as
contingent to the subjective construction process.
9 Cf. Kincheloe (1991, p. 26).

10 For a further discussion see e.g. von Foerster and von Glasersfeld (1999) and Rusch (1999).
11 Cf. Rusch (1999, p. 9).
Altogether the separation of knowledge and reality and the substitution of realis-
tic stances of knowledge verification with more context dependent operative validity
introduce elements of cognitive relativity. As truth cannot be defined in an absolute
sense, there are no objective criteria for the validation of constructs. However, in
a constructivist perspective the possibility of evaluating the resemblance between
constructs and reality is ruled out. This means that there are no objective criteria to
appreciate the degree to which constructs mirror reality and therefore to appoint a
certain construct as “true” or “false.” Instead, constructivism posits that constructs
can be evaluated according to their capability of granting human survival, i.e. ac-
cording to their “viability.”12
The category of “viability” can be applied to constructs concerning both senso-
rial perception and intellectual elaborations. In particular, a construct concerning
sensorial perception is said to be viable if it supports the individual in her (inten-
tional) interaction with the external environment, granting therefore the individual’s
survival given the state of the environment the individual is confronted with. Con-
structs that regard intellectual elaborations, strive for gaining knowledge and involve
speculation and abstraction, will be assessed as viable if what they conceive does
not contradict other viable constructs.13
Viability can be therefore profiled as a dynamic criterion which neither relies
on objective nor absolute standards. An operative definition of viability cannot
be formulated because viability is a highly contingent notion. It is based on the
contingency created by the concrete possibilities of actions of a certain individ-
ual, which are given by the actual states of the environment and the sensorial and
cognitive equilibrium of that individual. For this reason, viability accounts for the
provisory and individual validation of constructs. Constructs mirror their own gen-
esis. They reflect the feature of the cognitive process from which they origin. They
are the sensorial or intellectual product of a certain individual facing a specific
situation in a certain time and as such they also reflect the individual experience and
social history.14
The substitution of the objective truth-based validation criterion for constructs
with the dynamic criterion of viability allows for the possible heterogeneity of con-
structs, whose implications will be further discussed at Sect. 3.1.2 of Chap. 3.
However, underlining the subjectivity of constructs and admitting their possi-
ble heterogeneity, does not lead to solipsismus. Human knowledge, although it is
not permitted to penetrate reality, is vital to individuals because it supports their
behaviour and enables their survival. Despite the subjectivity of constructs, their in-
terpersonal validation of constructs can occur if they are accepted as viable among
a certain community. The interpersonal validation of a construct is produced and
enforced by the interaction between individualsand leads to the social construction
12 The concept of “viability” assumes a central importance for the constructivist approach. Cf.
e.g. von Glasersfeld (1995), Maturana and Varela (1987) and Rusch (1999).
13 Cf. von Glasersfeld (1995, p. 122).
14 More on viability will be discussed at Sect. 3.1.2 of Chap. 3.
of reality. Such a shared system of viable constructs creates a sort of “objectiv-

ity,” i.e. allows for the individuals belonging to that community to be in cognitive
equilibrium.15
3.1.1 Observer, Observation and the Construct of the Self
Individuals do not have a direct immediate access to reality. Their approach to it

can only be filtered by their own perception, which informs their way of cognitively
processing it. Therefore, cognition and knowledge do not represent acts toward the
discovery of reality, but toward the construction of cognitive reality.
Individuals cannot know reality, but only the observation they make out of it.
The differences between phenomena, objects or events, as they are perceived by an
observer, only play in their own minds. Human constructs inevitably carry and re-
flect elements of the subjectivity of the individual who formulates them. Constructs
are therefore relative and related to their subject and mirror the subject’s biological
determination, experience and social history.
It follows that observations cannot be abstracted from their observer, since they
only exist and assume their sense of being within the observer’s cognitive reality, in
which the observer is embedded and to which she belongs as well.
Radical constructivism differentiates between the concepts of “internal” and “ex-
ternal observer:” while the first applies to the individual whenever involved in her
own observation, the latter depicts the individual observing something external to
her own cognitive reality.
It is interesting that even as an internal observer, the individual does not gain
any privileged ontological stance. She can just process a construct of the “self,” but
cannot have any immediate knowledge of herself, i.e. not filtered through her own
cognition.
The construct of the “self” assumes a central role for the coordination and evalua-
tion of the individual’s behaviour. This is because the self-reflection of an individual
is the basis for the development of consciousness, which enables the perception of
the individual’s own behaviour as such and permits the individual to define cause-
effect patterns between her own actions and their consequences.
The human brain is interpreted as a subsystem which is steadily involved in the
reception and interpretation of external stimuli. Cognitive neuroscience describes
perception in a way which corroborates this view. It has been proved that external
stimuli hit the sensorial receptors and get communicated to the brain in the form
of electrochemical signals. All of these signals are qualitatively equivalent, inde-
pendent from which sensorial receptor they come from.16 They reveal the same
amplitude and only differ in their frequency, which expresses their intensity.17 This
15 Cognitive equilibrium is not an eternal (durable) state, but is contingent on the particular state
of the environment which is faced.

16 Cf. Lehmann-Waffenschmidt (2002, p. 23).
17 Cf. von Foerster (1992, p. 56 ff).
indicates, that even in a physiological perspective, perception can be characterised

as a process by which external stimuli are codified,18 whereas the brain does not
have any direct access to the entity from which such stimuli were generated. It can
be therefore said that the brain works in a self-referential and self-explanative way
in the sense that “alle Bewertungs-und Deutungskriterien muss das Gehirn aus sich
selbst entwickeln.”19 Memory assumes a central role for the perception process, as
it stores previous sensomotorical experiences and constitutes the key for their cog-
nitive evaluation.20
Other results of cognitive neuroscience support the constructivist interpretation
of human perception and cognition. They provide evidence that three different sets
of brain regions can be related to three different neuronal processes. A first set of
brain regions, namely the higher order sensory cortices, processes the perceptual
representation of stimuli. A second set, formed by the amygdaly, striatum, and or-
bitofrontal cortex, is responsible for the association of perceptual representation
with emotional response, cognitive processing and behavioural motivation. The
third set which consists of the higher cortical regions completes the construction
of an internal model of the social environment.21
The human brain has an undeniable physical presence, and can therefore be
analysed as any other external entity,22 that means individuals can observe it as ex-
ternal observers. There is an essential difference between the “real” and “cognitive”
world, i.e. between outward and inner reality. Individuals simply have access to the
cognitive world that they themselves create. Everything they perceive, all informa-
tion they process and every interaction they take part in, take place within the realm
of cognitive reality. Cognition acts as a filter and at the same time as the only link
between inner and outward reality, as it coordinates the processing of the stimuli
from the external reality.
Cognitive reality is closed in itself and it is only in its realm that the individual
experiences the difference between inner and external reality. The perception of the
individual’s own physical presence, of her own body, brain and brain activity can,
without exception, only be experienced and processed by the individual within the
bounds of her own cognition.
The cognitive reality an individual lives is the expression of a complex learning
process which builds from the interaction of the individual with the external envi-
ronment. This learning process can be configured as a continuous experience evalu-
ation, memorization and selection of successful behavioural patterns. The cognitive
process and specifically, the link between behaviour and cognition will be further
explored in the following paragraph.
18 Cf. Roth (1995, p. 59 ff).

19 “The brain has to develop all evaluative and interpretative criteria out of itself.” Author’s trans-
lation, cf. Schmidt (1987, p. 15).
20 Cf. Roth (1985a, p. 239).
21 Cf. Adolphs (2003, p. 166).
22 The brain as perceivable entity, that can be as such object of analysis, has been often used as
critical argument against constructivism. Roth’s argumentation (which follows) responds to such
criticism. Cf. Roth (1985a) and (1985b).
3.1.2 A Constructivist Approach to the Cognitive Processes
The framework of constructivism can provide a philosophical basis for deepening

the general mechanism through which individuals form their beliefs as well as an
explanation of their heterogeneity. As this approach merely allows for partial sub-
jective knowledge of the world, it takes into account both “the intrinsic limits of the
human mind in terms of computation and prediction capabilities. . . [and] the het-
erogeneity of agents in their beliefs and information endowments;”23 this is there-
fore coherent with the bounded rationality approach. Relying on the constructivist
approach, it seems possible to outline a framework in which the heterogeneity of
knowledge and information and the subjective rationality of the individuals can be
modelled in a mutually consistent way.
“Knowledge” should be distinguished from “information”; the former being a
“map from action to consequences. . . [which] is activated whenever the system
changes its state”24 the latter being the identification of a given state.25
In a constructivist way,26 “(i) knowledge is the output of active elaboration of the
subject ranging from the selection of external inputs to the constructions of “models
of the world”; (ii) the subject is continually engaged in the empirical control of such
models, which thus act as a feedback mechanism in the construction process.”27 This
implies, in particular, that the individual “models of the world” cannot be isomor-
phic with the external world. Also according to the bounded rationality approach,
cognitive activity produces stylized subjective mental models to support the indi-
vidual in her decision-making. The individual will then evaluate the feedback such
models receive from the external environment and, if satisfied, she will rely further
on those models; if not, she will modify them.
The cognitive process can be depicted as an input-output process with a feedback
mechanism (cf. Fig. 3.1):
External environment Internal environment
INPUT heuristic faculties OUTPUT

Validation
Empirical control and feedback
Fig. 3.1 The cognitive process in the constructivist approach (Tamborini, 1997)
23 Cf. Tamborini (1997, p. 255).

24 Cf. Tamborini (1997, p. 257).
25 Idem.
26 Here is meant the approach of Watzlawick (1981).
27 Cf. Tamborini (1997, p. 257).
Stimuli from the external environment, interpreted here in the form of physical
signals without any prior ontological assumption, are the inputs of the cognitive
process and are elaborated in the internal environment. The internal environment,
i.e. the “mind,” includes all thinking faculties.
To represent its functioning, the computational approach28 will be followed. Such
an approach focuses on the mapping from external to internal states, whilst the latter
can be interpreted as logical ordering of syntactic elements, “like steps in a computer
programme.”29
An alternative approach could have been the “neural approach,” which focuses
on the physical disposition of the internal environment, thus on the particular con-
figuration of the neuronal networks.30 While the neuronal approach seems to have
a better explanatory power for the unconscious mental processes, the computa-
tional approach suits the representation of conscious thinking and decision-making
better.31
According to the computational approach, the mapping from external to internal
states is based on a series of heuristic faculties, among which abstraction and causa-
tion are essential for rational decision-making. The internal environment produces
as output a representation or a mental model of the world, i.e. “a set of causally
ordered relationships [. . .] among selected objects or events, aimed at explanation
and prediction.”32 Causal ordering, i.e. how the human mind creates an efficient
order for action,33 plays a central role in human explaining and predicting. Ratio-
nal actions are based on causal models that forecast in the individual’s mind the
consequences of her actions.
Consistently, the bounded rationality approach also relies on the notion of mental
models. To elaborate her mental models, the individual selects the external signals
and combines them according to pre-existing patterns of configuration. The combi-
nations of signals have then to match with such patterns in order to be recognized.34
This process of selective perception can be called “abstraction” and regards both
physiological (e.g. of an object) and conceptual perception (e.g. of an immaterial
object or of a social situation).
The last stage of the cognitive process is represented by the validation of the
mental models. In the spirit of constructivism, this validation does not require an
internal (mental) representation to be an exact reproduction of the external reality
because this would just be a metaphysical ideal, deprived of any operational content.
Rather, the individual simply needs a rule that establishes that a certain model is
provisionally “valid” or “viable” for action.
28 Cf. Newell and Simon (1972), Simon (1977, 1981).

29 Cf. Tamborini (1997, p. 258).
30 Cf. McClelland and Rumelhart (1986).
31 Cf. Tamborini (1997, p. 258).
32 Tamborini (1997, p. 259).
33 Cf. Lorenz (1973).
34 Idem.
The constructivist approach introduces the notion of “cognitive equilibrium” as

a sort of measure to which the viability of a mental representation can be related.
An individual can be said to be in “cognitive equilibrium” if the actions generated
by her internal environment are consistent with her objectives, given the responses
from the external environment.35
To make the concept of provisional viability more operative, the notion of cogni-
tive equilibrium could be related to that of satisficing, since it allows that different
mind constructions can coexist and meet the subjective aspiration levels. Viabil-
ity is based on what the individual experience, whereas experience here should not
be interpreted as observation of events, but as action, because it is through action
that the individual tests her mental models. The result of such testing operates as
feedback on the construction of knowledge, which leads either to the validation of
actual mental models or to their modification. In this way, knowledge is not directly
a representation of the world, but a representation of the experience of the world.36
This feedback mechanism lets one characterise the cognitive process as self-
referential, because every construct, once confronted with the subjective experience
of the external environment, shall be reflected to the mind which originated it, i.e. it
shall be self-reflected. Human cognition is recursively engaged in the elaboration of
mental models out of external stimuli and in their evaluation according to the indi-
vidual experience of the external world, either to consolidate a (subjective) viable
construct or to modify a non-viable one.
Knowledge in the constructivist view can be characterised through the attributes
of partiality and possibility, as opposed to the objectivist ideal features of complete-
ness and necessity. This allows for heterogeneity and coexistence of mental mod-
els.37 Partiality stems from the conception of the individual as purpose oriented and
as such is guided by “interest,” i.e. any purpose (in a broad sense) that can motivate
(“cause”) the individual’s action. Thus interests “(i) elicit agents’ action and (ii) di-
rect agents’ heuristic procedures in construing an [intentionally] valid model of the
external environment. In cognitive terms, interests provide the focus for ‘conscious
devices’ aimed at reducing complexity through pattern creation and signal-pattern
matching.”38 Interest does not only provide a motivation for action, but can also
direct the cognitive process on which rational action relies. An immediate conse-
quence is that no one needs more knowledge than what she needs to manage the
situations she is usually confronted with; another one is that, since no isomorphic
representation of the external environment is possible, the selection of an absolutely
valid rule can be excluded. So, partial knowledge is the intentional result of the cog-
nitive activity of an interested individual and not just an exogenous constraint. As
such, it explains the persistence of heterogeneous mental models and the consequent
behavioural patterns.
35 Idem.
36 Cf. Maturana and Varela (1987).
37 Cf. Tamborini (1997, p. 261 ff).
38 Cf. Tamborini (1997, p. 261).
Since constructivist knowledge is insolubly connected to a particular experience,

it gives life to a constellation of different mental models. Their convergence to col-
lective shared mental models may take place or may not. If individuals interact for
a sufficient time, their possible (heterogeneous) mental models may eventually con-
verge to a common one (with common knowledge of this). However, there is also
evidence39 that, if the exchange of information is imperfect, the convergence of be-
liefs toward a common mental model may be excluded.
Though admitting the possible heterogeneity of beliefs and mental models, con-
structivism assumes isomorphism of the human mind; this means individuals do
not differ in the way they know. Also, isomorphism can be integrated with the
bounded rational approach: individuals assume isomorphism or symmetry between
themselves and others and they use the tool of introspection to form beliefs about
others.
3.1.3 Scientific Research and Reflexivity
The constructivist stance to knowledge contrasts the objectivist and rationalist con-
ception of a world which exists independently of human involvement in it and which
can be known in its constitutive objective features apart from the human knowing
activity. Constructivism challenges such a view in that it underlines how research
constructs what it claims to analyse: what is described by research studies does not
exist independently of such studies and their involvement with their own objects of
analysis.
Aside from an ontological perspective, it is here sustained that all that is known
can only be claimed to be known as a construct, i.e. as a model or a modelling
process that can be neither abstracted from the constructing process from which it
originates nor related to an external independent reality. The natural system and its
processes cannot be investigated without dealing (though without interfering) with
them and can only be known through the categories and distinctions made by an
observers’ community. This relates to the mechanism by which every distinction
insolubly links its components, as no outside can be defined without its boundary.
The research process should therefore be seen as socially constructing a world,
researchers being an integral part of it, as illustrated by the following statement: “As
inquirers and researchers, we create worlds through the questions that we ask cou-
pled with what we and others regard as reasonable responses to our questions.”40
In this perspective, reflexivity is an inescapable feature of research. It and cannot
be avoided and constitutes a source of distortion that affects different steps of the
research process. The awareness of it and of its implications calibrates the validity of
scientific explanation, in the sense that science cannot transcend the limits of human
39 Cf. Geanakoplos (1989).

40 Cf. Steier (1991a, p. 1).
phenomenon
translation
mutual
and stylized reformulation
mirroring
pontification
invention
abstraction of conditions and
intervention
observation
explanation
Fig. 3.2 The Reflexivity of Research (author’s representation)
constructing. Its validation can therefore be more realistically interpreted relying on

the relative and contingent category of viability.
Research aims at explaining phenomena and observation constitutes its first step.
The process of scientific explanation is typically articulated by explaining social
reality which means at first to observe a phenomenon and then to codify it in a styl-
ized way (see Fig. 3.2). This “stylized reformulation”41 is the expression of the cur-
rent mainstream codification of the situation in question. Such reformulation gives
evidence of some “abstract conditions” that can properly depict the situation in a
standardized and comparable manner, so that a suitable “theoretical framework” –
a theory or a model - may be found or worked out and an “explanation” of the
phenomenon can be achieved.
In formulating a phenomenon in a stylized way, researchers inevitably shape
what they are going to investigate, by means of assessing which features have to be
underlined as relevant, which ones have to be intensified by the analysis and which
ones can be neglected. This kind of construct of an observed phenomenon reflects
its creator (as all constructs do). Because of the mutual reflection process estab-
lished between observer and observation, research can therefore be characterised as
mutually mirroring.42 The self-referring character of mirroring can be appreciated
as a dialectical device: if the individual is deprived of the possibility of penetrating
reality, she can merely perceive it in antithesis to herself, which means that every
41 “This stylized reformulation is the actual common sense understanding of the scientific com-
munity of the real situation which gives the subject of the analysis.” Cf. Lehmann-Waffenschmidt
(1996, p. 46).
42 Cf. Steier (1991b, p. 173).
discourse about reality cannot transcend the limits of her understanding. The indi-
vidual is therefore mirrored by her own construction, which represents an outward
view of the individual and bends back, reflects the individual’s features.
When research focuses on social facts, mutual mirroring “works both ways when
dealing with human/social systems whose members are capable of creating meaning
and context for those situations in which they are embedded.”43 The members of a
social system construct the situations they are confronted with and at the same time
intervene and shape the course of the events they are participating in by means of
their constructs. The researchers’ constructs overlap with the members’ ones and
may influence each other mutually. It can also come to parallel processes44 between
researchers’ and social group members’ constructions, in the sense that researchers
are not immune to the social dynamics they investigate.
A scientific observation is based on specific abstract conditions which rely on the
standardized notation which is commonly shared by a certain scientific community.
This standardized notation includes categories, definitions and measures that “have
been built up and make sense based on the kinds of distinctions felt necessary for
activities of a particular research community.”45 In this sense, research resembles
a work of translation as well as a reformulation of perceived fact in the standard-
ised language of the research community. This holds both for the inquiry of natural
facts, whereby natural phenomena and patterns have to be linked with standard-
ised categories, and for the analysis of social facts. These are social constructs or
at least they reflect them. Translating is means for a constructionist researcher to
be “faced with the dilemma of wanting to understand ‘how others construct mean-
ing or make happenings’ (in their ‘language’), while recognising that he or she
is a member of community of particular researchers who have a particular lan-
guage with which they demand to be addressed, and that these two languages are
not the same.”46
Social research translates the constructs of a certain group into constructs of an-
other group, thus creating bridges, i.e. “pontificating,”47 between these two groups.
As every reference relation, the one established by social research develops between
the elements of a range and those of a domain. Each individual belongs to several
different social groups embedded in each other (the set of all those sets existing as
well), social scientists do not constitute an exception. Because of the non emptiness
of the intersection between range and domain, the reference relation which social re-
search establishes fulfils the requirements for being self-referential in a strict sense,
as it has been discussed at Sect. 1.2 of Chap. 1.
Research defines and classifies phenomena. Creating an order is a way to con-
struct new features of reality, which whenever perceived, can affect the course of
43 Cf. Steier (1991b, p. 173).

44 Cf. Smith and Crandall (1984) and Smith, Simmon, and Thames (1989).
45 Cf. Steier (1991b, p. 175).
46 Cf. Steier (1991b, p. 175).
47 Cf. Steier (1991b, p. 175).
the system observed. Analysis, as a way of interpreting phenomena, constructs its

object and contributes to shape them.
Because of its reflexive character, scientific research can be interpreted in an eco-
logical sense, whereas “ecology” here stands for “a context constituted by a fitting
together of ideas, ideas that here include a researcher (co-) constructing (with recip-
rocators) a world.”48 Mirroring indicates that a researcher should be located in such
an eco-system.
3.1.4 Science as Language Game
When focussing on science as a human construct, specifically on the involvement of

a researcher in the system which constitutes the object of her research as well as on
the similarities between research and translation, science can be characterized as a
language game.
It is just through the individual’s presence that words and sentences which build
a language assume meaning. Language is a human construction. As every construct,
it depends on the subject who formulates it and reflects some elements of the sub-
jectivity of its creator.
Consensus regarding the meaning of a certain sentence requires an interpersonal
coincidence of interpretation of the words it is composed of, as well as the linguistic
structures involved. Such coincidence develops among members of the same lin-
guistic community, as they take part in a continuous interaction process. This is a
process, which articulates over both linguistic and non-linguistic communication
forms and puts the basis of a complex learning process, which aims at granting indi-
vidual survival in social interactions. As a result, it is then possible to define certain
“language games”49 which are specific to a community and depict simple forms of
language “consisting of language and the actions into which it is woven.”50
Individuals speaking with each other are not exchanging meanings or informa-
tion but merely words and texts. Each of the individuals involved in the conversation
will then associate words with mental constructs and representations, though the
meaning of such words are assumed in their minds.51 This is essentially a subjective
operation, which cannot transcend its subjective character even among individuals
belonging to the same linguistic community. This is due to the fact that their expe-
rience and behavioural repertoire might differ throughout.52
Language as a means of communication assumes a central role for science, as
science relies on it to be expressed and transmitted. Typically, scientific texts (both
48 Cf. Steier (1991b, p. 180), referring to the interpretation of Bateson (1972, 1991).
51 Cf. Schwegler (1999, p. 18).
52 “Was unsere Partner verstehen, [. . .] kann sich nur in den Bedeutungen verwirklichen, die sie
augrund ihrer Erfahrung mit den Klangbildern der Wörter verknüpfen, die wir gebrauchen – und
ihre Erfahrung ist nie identisch mit der unsrigen.” Cf. Von Glasersfeld (1996, p. 92).
written and oral) take the form of descriptions or observations of phenomena in the
way they are perceived.53
Science is based on a specific methodology, which codifies how observations
should be made and which parameters and procedures should be respected. Such
operative criteria direct the interpretation of scientific statements and texts, because
it restricts the possibility of associating words and/or symbols with meanings. How-
ever, the definition of such operative criteria does not completely eliminate interper-
sonal degrees of freedom, as the assignation of meaning involves subjective mental
processes, according to which the scientific description is perceived as a phenom-
enon and is processed. The scientific description will be faced with individual men-
tal models and with already existing viable constructs, so that individual experience
and social history are relevant for the process of understanding, interpreting and
eventually accepting scientific statements and texts.
3.1.5 Constructivism and Economics
Constructivism represents a useful framework for analysing economics. On the one

hand this is due to the fact that economic models and theories are essentially con-
structs and can therefore be analysed according to the observer-observation scheme.
On the other hand constructivism applies to the analysis of economics because of
the pronounced self-referring character economic theories and models reveal which
indicates the autopoietic character of the economic systems and the possible causal
role of theories on the behaviour of the economic actors.
Economic models and theories are constructs that order phenomena and events
and typically try to establish causal relations among them. A creative operation is
to set a causal relation, because it is based on some assumptions of contingency,
which, if accepted by the actors belonging to the economic system, can influence
the dynamic of the system described. Individuals who face economic problems act
intentionally and inform their behaviour to their own mental representations and
models. The construction of causality in economics therefore represents a particu-
larly sensitive field because accepted theories are integrated with the individuals’
mental models and affect their behaviour.
Causality between events is articulated over the continuum between the two ex-
tremes of perfect causality and of full casualty. Perfect causality between events,
E1 , . . . En , implies that the events are linked and follow each other in a determinis-
tic way, so that from a certain event Ei the whole sequence of succeeding events
Ei → E j , j > i, can be determined. On the contrary, full casualty can be spoken
of between events if the events are independent from each other and succeed each
other in a random way.54
53 Cf. Schwegler (1999, p. 20).

54 Cf. Lehmann-Waffenschmidt (2006a, p. 6).
3.2 Recursivity of Social Theorizing and Predictability of Social Reality 67
The “post-hoc-ergo-propter-hoc” fallacy55 occurs when a coincidental correla-

tion is falsely interpreted as causality. Causality between events can be inferred in
a fallacious way if it is assumed that the preceding event must be the cause of its
successor, i.e. that given the sequence Ei → E j , j > i, Ei causes E j . If a fallacious
relation of causality is believed by the individuals, they will inform their behaviour
to it and might make it become true. “If men define things as real, they are real in
their consequences.”56
Economic thought further implies constructivist elements whenever it comes to
relate actual economic behaviour and decisions with subjective variables as inten-
tions, purposes and expectations of the decision makers. This is for instance the case
of management and organization theory, as well as marketing, finance and game
theory.57
Elements of constructivism can be also applied for the analysis of strategic think-
ing. For example, the strategic repertoire of individuals belongs the so-called “strate-
gic thinking of higher order.” An individual tries to anticipate the behaviour of her
counterparts, putting herself in the others’ shoes and reflecting on what she thinks
the others think she thinks etc. The ability of forming hypotheses and construct men-
tal models about the beliefs and the intentions of the others, considering that they
can differ from their own, is something potentially innate in the individuals.58 This
ability is depicted in cognitive psychology as “theory of mind” (tom).59
Tom implies that individuals who think about others’ beliefs and mental con-
structs reflect the constructs that they assume to be held by their counterparts and
insert them in their own constructs and mental representations. The supposed con-
structs of the others will therefore contribute in shaping the individual’s perception
of reality.
3.2 Recursivity of Social Theorizing and Predictability

of Social Reality
Social theorizing can be affected by recursivity in two different ways: first, a social
scientist inevitably is part of the system she analyses, and second, theories can affect
the state and the evolution of the social system they aim to describe.
Though the first sort of recursivity regards all forms of human theorizing (as the
scientist can never be completely disentangled from the reality she examines),60 the
55 Cf. Fulda, Lehmann-Waffenschmidt and Schwerin (1998, p. 366).

56 Cf. Thomas and Thomas (1928).
57 Cf. Lehmann-Waffenschmidt (2006a, p. 3).
58 Several studies provided evidence that this ability develops in 3 or 4 years old children. Cf.
e.g. Wimmer and Perner (1983).

59 For a discussion of “tom” in philosophy, see e.g. Davidson (2001) and Dennett (1987) and for
its application to cognitive psychology see e.g. Carey (1985) and Wellman (1990).
60 Consider for example the indeterminism problem for quantum physics.
fact that theories may interfere with the dynamic of the system they aim to describe
represents a peculiarity of the social sciences.
Reflexivity of this kind is implied by the so-called “observer-observation prob-
lematic,” and stems from the coincidence between subject and object of the analysis.
It therefore does not find a real equivalent in the natural sciences, as it can be asso-
ciated with the human analysis of certain natural facts rather than with the analysed
facts per se.
The most similar problem that can be discussed in this insight is represented by’s
uncertainty principle.61 This principle, whose most popular version is the one of
Schrödinger’s cat,62 emerges as a consequence of wave-particle duality and states
a trade-off between the accuracy of paired measurements of canonical conjugate
quantities. The accuracy of the first of the paired measurements is taken at the
expense of the accuracy of the second measurement within a mathematically pre-
dictable quantitative range. This yields for the reflexivity of the act of measurement
because it introduces an irreducible uncertainty, i.e. it affects the system it refers
to. There is however an essential difference which remains between the reflexivity
implied by the Heisenberg’s principle and the one which affects the social sciences.
This consists of the fact that Heisenberg’s uncertainty principle only arises in con-
comitance with and because of the human act of measuring and can, in the end, be
traced back to the observer effect. The uncertainty relation does not affect the course
of the entities per se, but it only affects the possibility of their measurements. While
in Heisenberg’s relation reflexivity is due to the presence of an external observer, in
the social sciences reflexivity can be ascribed to the individuals both in quality of
observers and of object of observation.
3.2.1 Social Predictions
Humans have always strived for a reliable way of divining the future, because being
able to anticipate future events is an important capability which can increase the
chances of human survival and improve the adaptation to the environment.
In the antiquity predicting the future was mostly related to religion. The key
to anticipating the future was represented by the connection with the divine. The
method consisted of the interpretation of natural signs. That was, for example, the
case of augurs investigating animal entrails or “reading” the bird’s flight. Some
equivalent manifestations however still survive today, e.g. gipsies reading the fu-
ture on a crystal ball, or future-tellers reading the tarot.
There are essentially two ways of predicting the future on a scientific basis,
namely either an empirical or a theoretic approach can be adopted. Morgenstern
(1972)63 illustrates the two approaches using weather forecasting as an example.
61 Cf. Heisenberg (1927, 2000).

62 Cf. Schrödinger (1935).
63 Cf. Morgenstern (1972, p. 705), who quotes Granger and Morgenstern (1970, p. 5).
The empirical approach consists of registering the weather map of a certain day
(e.g. today), indicating several parameters for example air pressure, temperature,
wind direction and speed, then searching among previous weather maps for a similar
one. The weather map of the day after will be then used as prediction for the coming
day. This method is “strictly empirical in the sense that it requires no theory, i.e., no
understanding of the mechanism which produces weather changes. It merely makes
the assumption that there exists some correlation of the weather at time t and at
time t + 1.”64 The limits of the empirical approach can be found in the huge amount
of data it requires and in the difficulty of finding an almost perfect duplicate of the
actual situation in the past.
The second approach to prediction overcomes this shortcoming and is more op-
erative. It is the one which is typically applied in scientific predicting. It consists
of inserting the data, e.g. today’s weather map, in a mathematical model which ex-
presses meteorological regularities in the form of equations. The solution of the
mathematical model represents the prediction of tomorrow’s weather. Approaching
prediction in a theoretical way the accuracy of the forecast depends on the theoreti-
cal knowledge in possession and the possibility of collecting precise data.65
3.2.2 Explaining and Predicting the Social Reality
Explaining and predicting natural phenomena can be seen as symmetric processes,

as the natural course is not affected by human perception and knowledge (“Nature
does not care – so we assume – if we penetrate her secrets”66 ). The possible reflex-
ive implications of knowledge constitute, as already discussed, a peculiarity of the
social sciences.
Explaining is the codification of a particular real situation or event by means
of abstracting some conditions and theoretical laws that apply. Predicting aims to
proceed in the opposite logical direction, i.e. starting from the observation of the
occurrence of certain conditions in the particular situation examined. By means of
application of the theoretical laws, which suit the conditions, the development of the
observed situation will be elaborated.
The common methodological ground between explaining and predicting can be
found in the explicating process, the process which leads to a theoretical substitute
for a pre-theoretical concept.67
As represented in the flow chart in Fig. 3.3, even if the social and natural sci-
ences are both based on the scientific methodology of explaining and predicting,
reflexivity disturbs the symmetry of those processes for social theorizing.
64 Idem.
65 Cf. Morgenstern (1972, p. 705).
67 Cf. Güth and Kliemt (2001, p. 1). For a definition of “explication”, cf. Carnap (1956).
SOCIAL REALITY self-referentiality
published unpublished
PREDICTION
(stylized formulation
stylized reformulation of a future state)
application of (forecasting)
abstraction of conditions
conditions
theoretical framework
EXPLANATION
Fig. 3.3 Explaining and predicting the social reality (author’s representation)
On the left side of the flow chart, the process of scientific explaining of social
reality is depicted, while the right side represents that of social predicting.
Explaining social reality means at first to observe a social phenomenon and then
to codify it in a stylized way. Anew, this “stylized reformulation” (“This stylized
reformulation is the actual common sense understanding of the scientific community
of the real situation which gives the subject of the analysis”) 68 is the expression of
the current mainstream codification of the situation in question. Such reformulation
gives evidence of some “abstract conditions” that can properly depict the situation in
a standardized and comparable manner, so that a suitable “theoretical framework” –
a theory or a model - may be found or worked out and an “explanation” of the social
phenomenon can be achieved.
Predicting social reality can be characterized as a successive step of the scientific
procedure, and be interpreted as a first test of the validity of scientific explaining.
For this reason, the process of social forecasting can be read as articulating in the op-
posite logical direction of that of explaining. Consequently, the forecasting process
goes top-down on the flow chart and starts from the “explanation.”
The first step in order to forecast a social phenomenon is to insert it (again) into
the “theoretical framework” which suited the “explanation”. The “abstract condi-
tions” that the phenomenon can be reduced to should be compared with those that
led to the choice of the theoretical framework. If they are coherent to each other,
68 Cf. Lehmann-Waffenschmidt (1996, p. 46).
a “stylized formulation of a future state” (i.e. a “prediction”) can be elaborated.

As already stated, it makes a great difference if a social prediction is published or
not. A published social prediction may lead social actors to modify their behaviour,
either in order to fulfil or to destroy the prediction’s content. In this way, social
predicting can influence its own object, i.e. the forecasted reality.
Every step of explaining and predicting the social reality can be affected by some
distortions, which mainly stem from the cognitive and computational limitations of
the individual. Going top-down, the theoretical framework can be affected by the
problem of the validity of the theories or of the models chosen; they may be mis-
specified or even ideologically distorted. Both the abstraction of conditions in the
explaining process and their application in the predicting process can, of course, be
affected by the application problem. Further, as every stylization is the reduction of
the redundant aspects of a complex reality, such an operation is in itself arbitrary.
Scientific stylizations, even if they are based on strong codified and standardized
methods, make no exception. Thus, both the stylized reformulation of a complex
real phenomenon (on which the explanation relies) and the stylized formulation
of a future state (the prediction) can only be selective and may be distorted and
undermined by information deficiencies.
A last sort of distortion is related only to the predicting process and it deals with
the publication of social predictions. Once again, if a prediction gets communicated
to social agents, it leads them to modify their behaviour. Whether the prediction gets
fulfilled or destroyed, the reaction of the social agents invalidates it on its essence.
Even a further prediction capturing the individuals’ reaction is condemned to be
similarly biased.
In general, social sciences face three fundamental problems concerning predic-
tions. 69 First, there is an essential trade-off between accuracy of a prediction and the
likelihood of it happening. A second problem concerns the necessity of assuming
the continuity of the regularities in which the prediction is based. The “continuity
assumption” requires that the situation, defined through several concrete conditions
for which a future development can be forecasted, will actually stay unchanged till
the predicted event occurs (or even not). Whereas the falsification of a prediction
does not prejudice the predicting activity per se, admitting the necessity of the con-
tinuity assumption points at one of the fundamental epistemological limitations of
predicting.
While these first two problems apply in an undifferentiated way to the pre-
dictability of natural and social facts because they are related to intrinsic limitations
of human predicting, a third problem exclusively refers to social forecasting. This
problem, which is known as the “Oedipus effect,” is represented by the reflexivity of
social predictions, which is responsible for the possible self-altering effect a social
prediction may have.
69 Tietzel refers to them as “Geltungs-, Anwendungs-und Ödipusproblem” that can be translated

as validity, applicability and Oedipus problem. Cf. Tietzel (1989, p. 546 ff).
3.2.3 Reflexive Predictions
A social prediction can act in a reflexive, self-altering way if its knowledge among
a certain community alters its predictive accuracy ceteris paribus. A reflexive pre-
diction is a prediction that fulfils the following conditions. First, the prediction has
to be revealed to the decision makers. It speaks in this sense of the “dissemination
status” of a prediction and can be distinguished between “made public” or “kept pri-
vate.”70 Second, in order for the self-altering effect not to be trivial, the prediction’s
informational content has to differ from the mental representations about the future
developments which are otherwise held by the individuals.71 Only in this case can
an eventual modification of behaviour be alleged to the prediction. A third condition
is that the prediction has to be believed by the decision makers.72
According to Buck (1963):
“A prediction is reflexive if and only if:
1. its truth-value would have been different had its dissemination status been
different;
2. the dissemination status it actually had was causally necessary for the social
actors involved to hold relevant and causally efficacious beliefs;
3. the prediction was, or if disseminated, would have been believed and acted upon;
and finally,
4. something about the dissemination status or its causal consequences was ab-
normal, or at the very least, unexpected by the predictor or by whoever calls it
reflexive, or by those to whose attention its reflexive character is called.”73
A further crucial point regarding the self-altering effect of revealed social predic-
tions is whether they are believed or not. The “compliance” with a prediction is tied
to the perception the agents have of its validity. For example, theoretically erroneous
statements could become self-fulfilling solely because they are believed by individ-
uals, or supposed to be believed by the majority of them. Similarly, the concept of
“sunspots equilibrium”74 refers in economics to situations in which market outcome
or allocation of resources depends on variables that only matter because individuals
believe they do.
The study of reflexive predictions can be traced back to the nineteenth century.
Venn (1966) discussed, for example, the concept of “suicidal prediction.” A vivid
sociological debate flourished on the reflexivity of Marx’s work, as illustrated by
the contributions of Merton (1936), Roshwald (1955) and Grünbaum (1956).75
70 Cf. Buck (1963).

71 Cf. Tietzel (1989, p. 554).
72 Idem.
73 Cf. Buck (1963, p. 361–362).
74 Cf. Cass and Shell (1983).
75 For a further interesting contribution of reflexive predictions and the possibility of economic
forecasting see Bombach (1962).

The fundamental sociological principle, known as “Thomas Theorem,” (“if men

define things as real, they are real in their consequences”76 ) explicitly refers to
the possible creative role conceptions and beliefs may have in shaping reality. This
implies that a situation can occur solely because of being conceived and illustrates
the mechanism on which reflexive predictions rely.
Merton (1936) interprets social public predictions as something which funda-
mentally biases the axiomatic construction they are built upon. For example, the
ceteris paribus condition, which is implicitly assumed by every forecasting, gets
inevitably violated because of the following prediction: “Public predictions of fu-
ture social developments are frequently not sustained because the prediction has
become a new element in the concrete situation, thus tending to change the initial
course of developments.”77 This gives account for the frequency and pervasiveness
of self-altering predictions. They typically perpetuate the “reign of error”78 they are
responsible for, because the occurrence of the predicted events apparently confirms
the validity of the prediction.
Reflexive predictions can either act self-fulfilling or self-destroying. A self-
fulfilling prophecy is “in the beginning, a false definition of the situation evoking
a new behaviour which makes the original false conception come true.”79 Self-
destroying prophecies are “suicidal” prophecies which do not survive their own
dissemination and acceptance. If the individuals are informed of the prediction and
believe in it, they will act which invalidates its predictive content.
A critical issue in estimating the degree to which a prediction is reflexive is rep-
resented by the counterfactual comparison between what actually occurred after
publishing the prediction and what would have occurred otherwise. The difficulties
that are associated with such an analysis are evident, because speculative tasks of
counterfactual analysis are involved.80
The possible causal effect of predictions on the predicted events has been related
by Popper to the legend of Oedipus. “Oedipus, in the legend, killed his father whom
he had never seen before; and this was the direct result of the prophecy which had
caused his father to abandon him.” Popper therefore suggested “the name ‘Oedipus
effect’ for the influence of the prediction on the predicted event (or, more generally,
for the influence of an item of information upon the situation to which the infor-
mation refers), whether this influence tends to bring about the predicted event, or
whether it tends to prevent it.”81 In Popper’s view, the reflexivity of social predict-
ing is so pervasive that potentially every social prediction can be invalidated by the
reactions it provokes among the social actors.
76 Cf. Thomas and Thomas (1928, p. 571–572).

77 Cf. Merton (1936, p. 903–904).
78 Cf. Merton (1936, p. 477).
79 Cf. Merton (1936, p. 477).
80 Counterfactual analysis is enjoying in the last times certain favour in the historical literature.
See e.g. the contributions of Demandt (2001), Ferguson (1999) and Salewski (1999).
81 Cf. Popper (1957, p. 13).
In economics, the theme of reflexive predictions was first emphasized in 1928

by Morgenstern’s seminal essay on economic forecasting, which originated a vivid
debate in the economic literature.82
Morgenstern approached the problem of economic forecasting in an extremely
direct way. While economic theory aims at enabling accurate predictions, the pre-
dictive power of the theory represents the ultimate test for the theory’s validity. The
point is that even false theories may often reveal a great predictive power.83 Ac-
cording to Morgenstern, predictions are only possible on basis of general theorems
(i.e. “in the large”84 ). Therefore, they cannot deliver really useful information be-
cause they merely consist of observations of the essence of a general theorem as
it applies to concrete settings and confirms the regularities it states. More precise
and informative historically concrete predictions (“in the small”85 ) are not possible
in the rule, because “the data with which the economic forecaster must deal are of
such a nature as to make it certain that the prerequisites for adequate induction
must always be lacking.”86 In addition, “economic processes [. . .] are not charac-
terised by a degree of regularity sufficient to make their future course amenable to
forecast”87 and furthermore “forecasting in economics differs from forecasting in
all other sciences in the characteristic that, in economics, the very fact of forecast
leads to ‘anticipations’ which are bound to make the original forecast false.”88
By strict logical reasoning, Morgenstern89 came to the conclusion that social phe-
nomena cannot be foreseen, since revealed social predictions influence the analysed
system in a way that can in principle never be correctly evaluated. Morgenstern’s
argumentation is that every social prediction is followed by a behavioural adjust-
ment of the social actors. Even a reformulation of this prediction, which takes into
account this feedback, will be followed by another adjustment etc. This infinite re-
adjustment process is known as the “Morgenstern Paradox” and can be represented
as in Fig. 3.4.
Given a prediction (P1 ) about a social event, it is reasonable to assume that its
object will react to it. Such a reaction invalidates the original prediction. Assum-
ing that the reaction (R1 ) is known to the forecasters, a new prediction, P2 , which
takes into account R1 , should now be formulated. P2 will also generate a reaction
to itself, R2 , so that another prediction, P3 , will be necessary. From a purely logical
point of view, this infinite recursive process between prediction and reaction makes
it impossible to correctly deal with social predictions and also with the theme of
the self-referentiality of social theory. About the question of whether the Morgen-
stern Paradox also has an empirical and not merely a logical validity, a vivid debate
82 See e.g the contributions of Grunberg and Modigliani (1954) and Bosse (1957).
83 Even if fundamentally false, the Ptolemean theory could calculate eclipses accurately and some-
times even astrology guess. Cf. Morgenstern (1972, p. 705).
84 Cf. Morgenstern (1972).
85 Idem.
86 Cf. Margert (1929, p. 313–314).
87 Cf. Margert (1929, p. 313–314).
88 Idem.
89 Cf. Morgenstern (1928, 1935).
behavioural
adjustment
Prediction 1 Reaction 1
prediction´s
update
Prediction 2 Reaction 2
......... .........
Fig. 3.4 The Morgenstern process (Lehmann-Waffenschmidt, 1990)
flourished.90 It can be demonstrated however (both in a mathematical and a prag-

matic way) that the conclusions to which the Morgenstern Paradox leads, are not
sustained from the evidence.
As Grunberg and Modigliani (1954) demonstrate, the Morgenstern Process
may also converge to a limit point, which represents the correct prediction of
the phenomenon, so that, as in Lehmann-Waffenschmidt (1996), the Morgenstern
Process only refers to one of the two conceivable cases which does not admit a
finite solution.
A pragmatic solution of the Morgenstern Paradox can also be formulated, con-
sidering that in reality nobody can perform infinite reflection processes. Two sorts
of limitations occur: the bounded rationality of the subjects and natural restrictions.
Bounded rational subjects are not able to perform infinite steps of recursive rea-
soning and are aware that their counterparts won’t be able to do that, as well. This
is proved, for instance, by the experimental evidence from beauty contest interac-
tions: in those situations where individuals have to guess what the others are going
to choose, subjects usually perform only a few (two or three) reflection steps.
Natural restrictions refer to time restrictions (the real choice-making process can-
not take an infinite time horizon) as well as to cost restrictions (time has an opportu-
nity cost). This explains why even fully rational subjects could not perform infinite
reflection processes.
Keynes’ analysis of reflexivity essentially concerns the issues of beauty-contest-
like economic interactions, conventional expectations and animal spirits.91 Al-
though Keynes does not explicitly address the theme of reflexive predictions, his
90 Among others see Bosse (1957) and Grunberg and Modigliani (1954). For a review cf. Lehmann-
Waffenschmidt (1990).
91 More on Keynes’s approach to reflexivity in economics, cf. Mackinnon (2003).
analysis corroborates the idea of an infinite regress which is established between

the economic theories, the individuals’ own and their counterparts’ beliefs. In order
to take a decision the agents consider what they assume the others are going to do
and are aware that the others are engaged in similar reflections. The individuals are
therefore engaged in a potentially infinite regress between expectations, individual
possible choices and others’ reactions. In particular, the notion of “conventional ex-
pectations” underlines the creative and manipulating power of the “prevailing view”
on individual behaviour, which can on this be basis assimilated to a self-fulfilling
prophecy, as it reveals a self-enforcing dynamic.
Plenty of examples of self-fulfilling as well as of self-destroying dynamics can
be mentioned. The disclosure of a public opinion survey can manipulate its results
by acting in a self-fulfilling way. Similarly, the German Federal High Court re-
cently had to decide on the responsibility of the Deutsche Bank for the bankruptcy
of Kirch’s corporation; Kirch accused the bank of having caused its bankruptcy
by publicly doubting its creditworthiness. Strikes warnings are typically associated
with the prediction of scarcity in the commodity supplied by the striking sector.
They sometimes provoke panic-stricken reactions among a population that make the
feared scarcity crisis become true and increase the prices. Similar hysteric reactions
occurred, for example, in Italy after the September 11, 2001 attacks in New York
and the US-attack on Afghanistan. As a consequence of media reports warning
against scarcity in the oil supply, which were not supported by real evidence, oil
prices rose.92
92 Cf. Lehmann-Waffenschmidt (2006a).

Chapter 4
On the Rationality of the Economic Actors
Modern economic theory follows the neoclassical approach and largely relies on
the paradigm of rational choice. The paradigm of rational choice interprets human
decision making as fully rational and based on Bayesian optimization of subjective
utility. This rational way of modelling human decision making stems from the ax-
iomatic characterization of utility and subjective probability rather than from the
direct empirical observation of the human economic behaviour.1 There is plenty of
evidence which questions the rational choice approach, both for its interpretation of
decision-making and for the assumptions on which it is based. Decisive critiques
also come from interdisciplinary studies which integrate economics with findings
from psychology, neurology, research on artificial intelligence and cognitive disci-
plines in general.2
From these criticisms alternative characterizations of the rationality of the eco-
nomic actors have been formulated, so that different “visions of rationality”3 coexist
in economic literature and assume different standards of rationality (see Fig. 4.1).
While models of full or perfect rationality assume that “the human mind has es-
sentially demonic or supernatural reasoning power,”4 models of bounded rational-
ity underline the boundaries of human reasoning. Perfect rationality of the economic
actors is interpreted as their capability of performing unbounded rational reasoning
and is modelled by probability theory and concretized in the optimization under
constraints.
Models of bounded rationality stem from the empirical observation of economic
decision-making and are essentially informed by the systematic violation of the
rational choice paradigm. Sub-optimal outcomes are not just odds that get elim-
inated because of the principle of survival-of-the-fittest. They configure in many
cases stable solutions. Admitting the boundaries of the subjective rationality means
1 Cf. Selten (1999).
2 For a critical approach to full rationality see e.g. Kahneman (2002), Gigerenzer and Selten (2001),
Güth and Kliemt (2004b), March (1994), Simon (1990, 1957). A more philosophical approach is
discussed e.g. in Kliemt (2001).
3 Cf. Gigerenzer and Todd (1999, p. 7).

78 4 On the Rationality of the Economic Actors
Standards of Rationality
Full Rationality Bounded Rationality
Satisficing Fast and Frugal

Heuristics
Fig. 4.1 Approaches to the rationality of the economic actors (author’s representation)
to interpret the deliberate decision-making process as “the ability to construct new

representations of problems”5 and stresses the “distinction between two types of
cognitive processes – the effortful process of deliberate reasoning on the one hand,
and the automatic process of unconscious intuition on the other.”6 As represented
in Fig. 4.1, there are essentially “two main forms of bounded rationality: satisficing
heuristics for searching through a sequence of available alternatives, and fast and
frugal heuristics that use very little information and computation to make a variety
of kinds of decisions.”7
This chapter offers an overview on the rationality debate in economics. After
discussing the unsolved dualism between rationality assumption and psychology
of choice, some of the solutions adduced in defence of the neoclassical paradigm
will be presented. The bounded rationality approach will then be introduced and
it will be focussed in particularly on the pioneering contributions of Simon and
of Kahneman and Tversky. The illustration of some principles of human problem-
solving, framing-effects and prospect theory will then conclude the chapter.
For convenience, the analysis of heuristics (both as sources of behavioural bi-
ases and as ecological aspects of bounded rationality) will be examined in the next
chapter. This will serve as an introductory discussion on the debiasing research. The
debiasing research will be considered as a possible inspiration for developing a suit-
able framework for the experimental analysis of the self-referentiality of economic
theories and theory absorption.
4.1 Questioning the Descriptive Validity of Rational Choice

Theory
The first criticisms of the theory of rational choice developed parallel to the dis-
covery of some of its seminal findings, as shown e.g. by the 1944’s Von Neumann-
Morgestern’s utility function and Allais’ experiments which question its validity.
The experiments were run shortly after.
5 Cf. Egidi (2005, p. 18).
6 Idem.
4.1 Questioning the Descriptive Validity of Rational Choice Theory 79
In the Fifties, the development of linear and dynamic programming enhanced the
interest towards rational choice theory. At the same time, however, the increasing
complexity of the models enforced the doubts on the plausibility of assuming real
agents to be able to perform in such a sophisticated way and to correctly solve
complicated analytical tasks.
A fundamental dilemma facing economics is constituted by the question whether
constrained maximization conveys a proper description of the economic decision-
making or whether its validity has to be rather confined to a normative level. At-
tempts towards the solution of this dilemma have been elaborated both by supporters
and by sceptics of the rational choice theory. While the former essentially defends
the neoclassical paradigm by stating the negligibility of the psychology of choice
when dealing with economic decision-making, the latter formulates on the basis of
the internal limitations of the subjective rationality alternative (more ecological and
less standardised) approaches to decision making.
4.1.1 The Neoclassical Defence
Initially, “utility, be it total or marginal, was considered a psychic reality, a sen-

sation that became evident from introspection, independent of any external obser-
vation [. . .] with directly measurable proportions.”8 In this sense, utility was not
considered to be an intrinsic quality of the alternatives faced by the actors, but
something dependent on their subjective evaluation. Its measurability and ordinal
rating represent therefore critical issues and even create problems which involve the
psychological mechanisms underlying the individual decision-making. Substituting
the notion of “utility” by that of “preference” constituted a device for separating
rational choice from its underlying psychology and represented an answer of the
neoclassical approach to the critique of its axiomatic fundaments.
The substitution of “utility” with the much more manageable concept of “pref-
erences”9 enhanced the construction of the axiomatic model of choice by assuming
the preferences’ completeness, transitivity, continuity and independence.10 In addi-
tion, it was attempted to anchor the theory of rational choice to empirical data ob-
serving the behaviour of large aggregates of individuals. For example, the S-shaped
expected utility curve11 was elaborated on this basis, whereas it did not resist the
test with data form representative samples.
The attempts of defending the neoclassical paradigm explicitly avoid the question
whether the individuals are in possession of the proper formal tools to adequately
optimise or not and strive toward the definition of new analytical instruments for
increasing the range of situations to which rational choice can be applied. All in all,
8 Cf. Schumpeter (1954), interpreting Menger and Böhm-Bawerk’s stance on utility.

9 This can be ascribed to Pareto.
10 Cf. Egidi (2005, p. 3).
11 Cf. Egidi (2005, p. 4), referring to Friedman and Savage (1952).
they fail to address the problem of the descriptive validity of the standard model of
rational behaviour at its core.
An exception in this sense is represented by Friedman’s defensive line of the
“as if” approach.12 Friedman’s main argument is that even if individuals are not
in possession of the proper analytical skills to correctly solve tasks of constrained
optimization, they behave “as if” they were. The conception underlying the “as if”
approach is that individuals might not be aware of the mental processes actually
involved in their decision making, as tacit knowledge plays a considerable role in
human choice making. Similarly to bikers keeping themselves in equilibrium with-
out being necessarily aware of the physical laws of motion, individuals confronted
with economic problems solve them in a rational way, without being aware of how
their decision emerged.
The “as if” approach further postulates that the individual preferences are not
observable because the individuals are not able to express them consciously and
explicitly. In this perspective the only variable that can be observed is the decision
itself, while the process of decision-making remains a black-box. It cannot be inves-
tigated because of the impossibility of empirically checking both the way decisions
emerge and the individual preferences. An additional problem is the unconscious-
ness of the individuals upon these variables. Their analysis is even not considered
to be of interest for economics, as selection ensures rational behaviour to be es-
tablished. Sub-optimal non-rational decisions will be gradually eliminated; agents
who do not perform constrained optimization correctly are punished and progres-
sively excluded. Because of the principle of survival-of-the-fittest optimal outcomes
eventually represent the only sustainable result of the economic decision-making.
Until today the “as if” approach has enjoyed a broad consensus among econo-
mists. It represents an elegant but fallacious attempt of justifying the axioms un-
derlying rational choice theory by stating the non reliability of the individual’s
expression of her own preferences as well as of non-consciousness of her own de-
cision emergence. Its way of simply avoiding the psychology of choice has been
seriously challenged by Allais’ and Simon’s criticisms, as it will be explained as
follows.
4.1.2 Allais’ Experiments
Running an experiment in which subjects were faced with pairs of binary choices,
Allais showed systematic violations of some of the axioms underlining the expected
utility theory.
In Allais’ experiment the participants were first asked to choose between the
following two alternatives:
(A) Certainty of receiving 100 million (francs)
(B) 10% chance of receiving 500 million, 89% of receiving 100 million and 1% of
earning zero
12 Cf. Friedman (1953).
4.1 Questioning the Descriptive Validity of Rational Choice Theory 81
Second, subjects had to choose between alternatives C and D:

(C) 11% chance of receiving 100 million and 89% of receiving zero
(D) 10% chance of receiving 500 million and 90% of earning nothing13
According to the expected utility theory, the individuals who prefer A to B
should prefer C to D as well, which is what was systematically violated in Allais’
experiment.
Allais’ results did not have the destabilizing effect that could have been expected
because they were qualified as referring to an extreme case which concerned a hy-
pothetical choice. Despite of the mainstream scepticism about Allais’ results, later
experiments, in which participants were paid, proved that the inconsistency in the
individuals’ preferences occurred systematically.14 The initial reaction to Allais’
research was that even more sophisticated versions of the expected utility theory
under uncertainty were formulated, trying to generalize its results and relax some of
its underlying axioms.
Allais’ results began to enjoy a broader acceptance by the mid-1970s, when they
became supported by many other relevant studies which seriously questioned the
descriptive validity of expected utility theory. Among them “the ubiquity of EU
[expected utility] violations in choices, process data, and elicitation procedures;
the elegance of Kahnemann and Tversky’s batch of new paradoxes; and Machina’s
(1982) assimilation of some of the empirical evidence against EU and introduction
of sophisticated tools for doing economic theory without the independence axiom”15
can be mentioned.
During the 1980s, further revisions of the expected utility theory coexisted with
the growing interest for alternative paradigms that insisted on the psychological
aspects of choice. Among the firsts, the weighted utility theory,16 the regret the-
ory17 and the disappointment theory18 can be mentioned, although it should be
noted that none of them received a statistical confirmation over their full domain
of applicability.19
4.1.3 Simon’s Bounded Rationality Approach
A decisive attack to the paradigm of rational choice came from the theory of
bounded rationality, which was developed by the research group lead by Simon at
the beginning of the 1950s. Simon and associates were enquiring administrative and
managerial behaviour. Their critique of rational choice builds on the unsuitability
13 Cf. Allais (1953, p. 527), from Roth (1995, p. 8).

14 Cf. Camerer (1995, p. 626).
15 Cf. Camerer (1995, p. 626). Cf. Camerer (1995) for the references, as well.
16 Cf. Chew and McCrimmon (1979).
17 Cf. Loomes and Sudgen (1982).
18 Cf. Gul (1991).
19 Cf. Egidi (2005, p. 7), referring to Hey (1991).
of this theory firstly, for describing decisional processes in organizations and sec-
ondly, for expanding to more general aspects of psychology of choice and decision
emergence.
The extreme complexity of constrained optimization immediately emerged as
an insurmountable limit for modelling the results of the field studies on economic
organizations conduced by Simon’s research group. In particular the analytical and
computational tools required for optimizing properly turned out to be unrealistically
demanding for individuals dealing with concrete decisional tasks in organizations.
Simon inferred from the observation that neither the individuals’ knowledge nor
their computational abilities allows them to optimize correctly, that both internal
(i.e. mental) and external (i.e. environmental) constraints limit human rationality.
He therefore characterized human rationality as essentially “bounded.”
Internal and external limits of the subjective rationality are linked to each other,
so that “human behaviour [. . .] is shaped by a scissors whose two blades are the
structure of the task environments and the computational capabilities of the ac-
tor.”20 The individuals have to take their decisions in a complex environment in
which it is very difficult and costly to explore and evaluate all available alternatives.
Additionally, individuals are also subject to time limits. Exploring the environment
constitutes therefore the first step of the decision process “through which the rele-
vant information is gathered and the appropriate knowledge is structured.”21 The
decision inspires to the results of the exploration of the environment and further de-
pends on the cognitive and computational tools the individuals are in possession of.
4.2 The Bounded Rational Revolution
The focus of Simon’s analysis lies on the notion of “subjective rationality,” in con-
traposition to the objective, absolute standard of rationality postulated by the neo-
classical theory. In Simon’s words, “in a broad sense rationality denotes a style of
behaviour that is appropriate to the achievement of given goals, within the limits
imposed by certain conditions and constraints. [. . .] The conditions and constraints
[. . .] may be perceived characteristics, or they may be characteristics of the organ-
ism itself that it takes as fixed and not subject to its own control. The line between
the first case and the other two is sometimes drawn by distinguishing objective ra-
tionality, on the one hand, from subjective or bounded rationality, on the other.”22
The special emphasis given to the notion of subjective rationality finds among its
most influential precursors Popper (1967) and points at the discrepancy between the
Olympic standard of full rationality and the behavioural patterns as they can be ob-
served in the reality. On this basis, the subjective logic and the reasoning processes
20 Cf. Simon (1990, p. 7).

21 Cf. Egidi (2005, p. 6).
22 Cf. Simon (1982, p. 8).
4.2 The Bounded Rational Revolution 83
which underlie individual decision-making are sought to be understood rather than

qualified as irrational or based on erroneous premises.
One of the most innovative contributions of the bounded rationality approach is
the attention it pays to the explorative phase of the decision-making process. In this
way, both the constructive activity of human reasoning and the influence subjective
mental representations have on the outcome of the decision process are in the fore-
ground. The bounded rationality approach does not restrict its critique to the rational
choice theory to the too demanding assumptions that are posited to the individual
computational abilities. Instead, it extends its criticism to the neoclassical hypoth-
esis viewing the individual problem agenda as exogenously given: “Neoclassical
economic theory assumes that problem agenda, the way in which problems are rep-
resented, the values to be achieved (utility function), and the alternatives available
for choice have all been given in advance. It has no systematic way of explaining
how problems get on the agenda, [. . .] what it is that people value and how values
change, or how action alternatives are created [. . .].”23 Trying to discover the logic
behind decision making reveals that only a “theory that deals with problem forma-
tion and with the design of solution alternatives can provide the basis for a theory
of economic change and development.”24
For investigating the mechanisms underlying human cognition, Simon developed
a new method of analysis, based on the proof principle according to which “if a
problem can be clearly described with appropriate language, then it can be trans-
ferred into a form computable for a machine.”25 The artificial reproduction of hu-
man thought can convey a picture of how individuals cognitively select and process
information in order to build the subjective frameworks on whose basis they will
then take their decisions.
This has been carried on by Simon e.g. through the observation of individuals
playing chess. The theoretical analysis of the game of chess and the observation of
the players’ behaviour was chosen as an ideal setting for investigating the individual
computational capabilities and their limits. In addition, the players’ verbalizations
on their own reasoning during the game were protocolled and helped the under-
standing of how decisions emerge.
Simon’s analysis reveals that individuals approach decision-making totally dif-
ferently than rational choice theory assumes they do. As illustrated by chess playing,
“chess strategies are inter-temporal decisions which require players to elaborate
and re-elaborate their analyses; their decisions are based on a process of learning
and mental model building repeatedly at odds with perfect rationality.”26 Decision-
making can therefore be characterized as an extremely complex process which in-
volves the interaction between the mental activities of induction, reasoning and
problem-solving. Problem-solving can, in particular, be depicted as central feature
of the individual mental processes.
23 Cf. Simon (1992, p. 5).

24 Idem.
25 Cf. Egidi (2005, p. 6).
26 Cf. Egidi (2005, p. 7).
A further feature of the individual approach to decision-making, which starts

to be profiled in Simon’s research, is that human decision-making is both strived
for by deliberate and non-deliberate mental processes. Obviously, using artificial
intelligence to simulate and reproduce human reasoning can solely capture the de-
liberate mental processes, i.e. those processes individuals are aware of and are able
to express by means of introspection. The decomposition of rationality in its consti-
tutive components of intuition and reasoning represents the core of Kahnemann and
Tversky’s research and will be discussed later in Sect. 4.3 of Chap. 4.
4.2.1 Adaptive and Satisficing Behaviour
In Simon’s words, “human reasoning, the product of bounded rationality, can be

characterised as selective search through large spaces of possibilities. The selec-
tivity of the search, hence its feasibility, is obtained by applying rules on thumb, or
heuristics, to determine what paths should be traced and what ones should be ig-
nored The search halts when a satisfactory solution has been found, almost always
long before all alternatives have been examined.”27
Bounded rational decision-making can be characterised as a creative process,
whose results depend on the paths followed by the individual in her exploration
of the environment and of the possible alternatives of choice. Because of subjec-
tive (i.e. related to the bounds of subjective cognition and computation) and objec-
tive (time and costs related) constraints, individuals are not likely to explore the
complete range of all feasible alternative choices. They would rather carry on their
search until they find a satisfactory alternative and then interrupt it. Therefore, “sat-
isficing” behaviour is based on a completely different mechanism than optimising
behaviour.
Optimising behaviour does not represent a feasible decisional procedure for
bounded rational individuals because it would require them to be much more sophis-
ticated and computationally skilled than what in fact they are. Constrained optimiza-
tion in particular necessitates the abilities to know the whole domain of all possible
outcomes, to asses the respective conditional probabilities and to rank them accord-
ing to the subjective preferences. In addition, each alternative has to be related with
its corresponding expected utility. This operation, together with the requirements of
full knowledge of outcomes domain, of complete ordinal preference structure and of
appropriate probability estimating is without any doubt too cognitively demanding
to provide a valid description of how human decisions emerge.
The bounded rationality approach decisively simplifies the abilities that are in-
volved in its description of the decisional process. It assumes the costly exploration
of the outcomes domain is interrupted as soon as the individuals evaluate an out-
come as “satisfactory,” i.e. as good enough to meet their subjective aspirations.
The satisficing approach to decision-making is therefore based on extremely simple
27 Cf. Simon (1992, p. 4).

dynamics, which does not involve complex analytical and computational capabili-
ties. Individuals behaving according to the satisficing principle have to solely rely
on a two-valued function because they simply have to evaluate the outcomes in a
binary way, i.e. as satisfactory or not.28
An individual behaves adaptively, if she adjusts her behaviour to extant condi-
tions in order to fit the current conditions of the environment she is confronted with.
In this sense, “to satisfice” means to adapt to the environmental conditions, i.e. to
reach an outcome that suffices the subjective aspirations. This underlines how the
process of decision-making is shaped by the individual perceptive and cognitive
capabilities on the one hand and by the structure of the environment on the other.
The notion of “environment” here does not apply to the description of the phys-
ical characteristics in the context in which the decision occurs, but rather encom-
passes all the aspects of the context which can be relevant for the individual and her
mental representation of the faced situation. The specification of the environment
will therefore “depend upon the ‘needs,’ ‘drives,’ or ‘goals’ of the organism, and
upon its perceptual apparatus.”29
For convenience, the implications of the satisficing rule can be effectively illus-
trated in a very simple setting: Consider a simple organism that has a single need
or goal, namely getting food, and has the choice between three possible activities,
which are resting, exploration and food getting.30 The environment which is faced
by this organism can be represented as a bare surface on which the organism can
move and over which food is randomly spread in one-meal piles. The organism’s
internal limitations are that the organism is not able to view the totality of the envi-
ronment at a glance, but just to a circular portion of it and that it has fixed maximum
rates of motion, of food storage and metabolizing.
This setting is particularly simple, because of the following features: the organ-
ism has the single goal of getting food, which does not compete with alternative
needs, its aspiration level remains unaffected by successes or failures and, while
food is randomly being distributed over the surface of motion, no searching pattern
is in principle better than the other, so that the organism does have to plan its search.
Finally, motion does not serve any other aim than reaching food.31
Because of the extreme simplicity of the setting and of the organism’s needs,
choice will be simple as well. In its search for food, the organism (a) explores the
environment at random and (b) whenever it sees a food pile it proceeds towards it
and eats it. Since it is able to rest, it is plausible to assume that “(c) if the total
consumption of energy during the average time required, per meal, for exploration
and food getting is less than the energy of the food consumed in the meal, it can
spend the remainder of its time in resting.”32
28 Cf. Simon (1955).

29 Cf. Simon (1956a, p. 40).
30 Cf. Simon (1956a, p. 41).
31 Idem.
32 Cf. Simon (1956a, pp. 41, 42).
According to the satisficing approach, even in more complex settings decision-

making articulates over the three processes of aspiration formation, satisficing and
aspiration adjustments.33 Individuals first set their aspirations, which can be in-
terpreted as discrete levels of goal achievement (process of aspiration formation).
Second, they try to fulfil such aspirations by using the means they possess which
they consider appropriate for the achievement of their goals (process of satisficing).
Third, they evaluate the feedback their choice receives and either confirm or adapt
their aspiration levels (process of aspiration adaptation or adjustment). Aspiration
levels can represent both “aspects of preferences and restrictions of decision alter-
natives, illustrating that in bounded rationality theory the separation of means and
ends, as known from the rational choice approach, may not apply.”34
Compared with the bounded rational approach, rational choice theory has the ad-
vantage of being an axiomatic elegant construction which can be easily applied in a
task transcending way. Once specified the individual structure of preferences,35 the
expected utility theory can be universally applied to predict the outcome of decision-
making in different settings. Despite the questionability of the predictive content of
rational theory, the advantage of its task of transcending applicability is what per-
haps motivated the resistance of mainstream economics to abandon such paradigm.
The satisficing approach and the bounded rationality theory in general do not
offer operative frameworks which can be applied in an undifferentiated way to a
variety of decision settings. Rather, “the satisficing approach so far offers only
an intuitive and natural terminology – already quite an achievement – but hardly
any specific guidance when trying to predict (economic) decision behaviour or to
give advice like in teaching and consulting.”36 Some recent experimental studies
are now discussing questioning the extent to which the satisficing approach is task
transcending.37
Similarly, the models of adaptive behaviour are extremely simple but highly task
specific. The plausibility of their way of modelling economic behaviour can be al-
leged to their simplicity, while their specificity hampers their undifferentiated ap-
plicability to a broad range of tasks.
Adaptive behaviour can be modelled on the basis of the psychological “Law
of Effect.”38 Given the individual initial propensity of choosing certain behaviours,
future choices will be influenced by already experienced responses. Behaviours both
in the form of simple actions and rules of thumb which could be associated with
positive outcomes will reinforce and be more likely to be repeated.39
33 Cf. Güth (2006, p. 1).

34 Cf. Güth (2006, p. 2).
35 Whereas considerations of the non-plausibility of explicability and completeness assumptions
still apply.
36 Cf. Güth (2006, p. 2).
37 Cf. e.g. Güth (2006) and Fellner, Güth, and Martin (2006a).
38 Cf. e.g. Bush and Mosteller (1955) and Roth and Erev (1995).
39 Cf. Stahl (1995, p. 304).
Behavioral Select Rule with Apply Rule to Observe

Rules Probability p Choose Action Outcome
Mutation
Update Rule Evaluate Rule
Innovation
Weights Performance
Imitation
Fig. 4.2 Simple model of adaptive behaviour (Stahl, 1995)
Figure 4.2 illustrates a simple model of adaptive behaviour. This is centred on

the evaluation of the feedback a certain behaviour yields for, to which the individual
rules are updated.
A consistent part of the research on bounded rationality can be interpreted as
an attempt of making simple adaptive behaviour models (such as the one presented
at Fig. 4.2) operative. By doing that, three fundamental questions are addressed.
“First, what is the empirically relevant class of rules? Second, what are the initial
propensities: Do all individuals have the same initial propensities, or is there sig-
nificant diversity among individuals? Third, what are the specific dynamics: how
is reinforcement quantified, how much weight is given to old information and new
information, how much experimentation/search is employed, how are new rules in-
troduced and how significant are computational errors?”40
In this light, the “heuristics and biases” program of Kahnemann and Tversky41
together with the “ABC” research program on “fast and frugal heuristics” lead by
Gigerenzer and Todd42 can be interpreted as striving toward the deepening of the
first question. The analysis of individual propensity, the measure of their determin-
ism as well as the factors influencing their formation is object of conspicuous re-
search on cognitive psychology.43 Research on aspiration formation and satisficing
emerged to answer the first part of the third question44 because they focus on the
mechanisms of feedback evaluation, while intertemporal mechanisms for weighting
information developed from the research programs on learning.45
4.2.2 Principles of Problem-Solving
One of the peculiarities of the bounded rational approach is to underline the central-
ity of problem-solving activities in individual behaving and decision-making.
40 Cf. Stahl (1995, pp. 304, 305).
41 Cf. e.g. Kahneman and Tversky (1974) and Kahneman, Slovic, and Tversky (1982).
42 Cf. e.g. Gigerenzer and Goldstein (1996) and Gigerenzer and Todd (1999).
43 For a review see Rabin (1998).
44 Cf. e.g. Güth (2000), Simon (1956a, 1956b, 1990), Todd and Miller (1999).
45 Cf. e.g. Camerer (2003a), Nagel (1995), Stahl (1995).
Human problem-solving is an essentially creative operation whose results

emerge as a discovery and cannot be therefore a priori predicted. They cannot
even really be simulated by reproducing the exact initial conditions under which
the problem-solving activity takes place. This adds to the difficulty of positively
modelling bounded rational behaviour and bounded rational decision emergence.
Some general features and principles of human problem-solving can be stylized
from the empirical and experimental investigation of the individual behaviour in
puzzles and games as well as in complex tasks like e.g. playing chess, mathemat-
ical problems, understanding instructions, detecting and inferring patterns between
numbered and numerable elements.46
Problem-solving involves the selective search through large spaces of possibil-
ities. The complete exploration of all the possibilities, which would be necessary
for an optimal outcome to be reached, requires too much time and is therefore not
feasible. In comparison with complete trial-and-error search, selective search has
the advantage that it enables the individual to find and settle on an outcome within
a reasonable time horizon.
Selective search is based on rules of thumb or heuristics, whereas heuristics are
simple rules that provide easy replicable methods for solving complex problems.47
Of course, “the more difficult the problem and the less efficient the heuristics, the
more search will be involved.”48 Certain heuristics can only be applied to specific
tasks, while others are quite general in their domain of application. Whenever exis-
tent and accessible to the individual knowledge, domain-specific heuristics will be
preferred to general ones, so that the individuals rely on them whenever they can.
Otherwise, they apply more general heuristics. These are also called “weak meth-
ods,” as they do not make specific use of the information about the specific domain
of the task to be solved. Individuals facing new problems, which typically demand
exploratory and creative skills, often have to rely on general heuristics. In all likeli-
hood, because of the novelty of the tasks, the individual cognitive repertoire does not
contain suitable specific heuristics and patterns of solution; as a result individuals
typically inform their problem-solving to weak methods.
Some bounded rational heuristics can be qualified as domain-specific or general,
depending on the amount of domain-specific knowledge which is available to the
individual. For example, the mean-ends analysis heuristic, which is one of the most
widely applied, always relies on the same mechanism of comparing between the
given initial state of the environment and the individual goals. For the individual
using memory cues is a possible means to reduce the gap between the given initial
state and the final individually striven configuration. The mean-ends analysis can
either be characterized as a weak or strong method, respective to either the indi-
vidual’s possession of little domain-specific knowledge or a conspicuous amount of
such knowledge.49
46 Cf. Simon (1988, p. 108).

47 Heuristics will be analysed in more details in Sects. 1 and 2 of Chap. 5.
48 Cf. Simon (1988, p. 108).
49 Cf. Simon (1988, p. 108 ff).
In Simon’s terminology, domain-specific knowledge is stored in the memory in

the form of “productions,” or “actions (A) paired to conditions (C).”50 If a problem-
solver recognises certain conditions to suit the situation she is confronted with, she
searches in her memory for similar conditions and then calls back the actions that
are associated with them. In this sense each “execution of a production is an act of
recognition.”51
Experts’ skills derive also from the large number of productions stored in the
individual memory because domain-specific knowledge is obviously larger among
those individuals that are acquainted with a certain problem. In this sense, individual
expertise can promote the rapid achievement of satisficing solutions. It should be
noted, that “in this picture of expertise, there is no sharp boundary between ‘insight’
or ‘intuition’ and analysis”52 both of which are “acts of recognition based on the
stored knowledge of the domain.”53
The experimental examination of the self-referentiality of economic theories, as
will be presented in Chap. 7, can be viewed as an attempt at testing whether adding
to the theoretical knowledge of the experimental subjects makes the number of pro-
ductions available increase and their recognition abilities improve.
Individuals make use of the three capabilities, insight, intuition and recognition
in different proportions, depending on the sort of the task to be solved: “for many
simple, everyday problems, recognition alone may be enough, and little analysis
may be necessary. In more complex situations, recognition of salient clues allows
analysis to take larger and more appropriate steps than if the heuristic search has
to depend on real methods alone.”54
4.2.3 Problem-Solving in Games and Puzzles
Some other interesting features of human problem-solving can be inferred by ob-

serving how individuals behave in games and try to solve puzzles.
Games are typically situations in which information is scattered, incomplete and
difficult to process.55 The individuals therefore face the challenge of collecting in-
formation, distinguishing relevant from superfluous information and trying to re-
duce the computational complexity associated with information processing.
The individual behaviour in games is strategic, although neither perfectly
forward-looking nor fully rational. Despite that, game theory assumes that given
the initial configuration of the game and its rules the agents can achieve perfect
knowledge of the game tree. This means that they are supposed to be able to infer
50 Cf. Simon (1988, p. 109).

51 Idem.
52 Cf. Simon (1988, p. 109).
53 Idem.
54 Cf. Simon (1988, p. 109).
55 Cf. Egidi (1992, p. 148).
all possible future configurations and developments of the game, to associate each
possible outcome with the corresponding pay-off and to assess the probability of
each outcome to be chosen. From the exploration of the game tree, a rational player
can therefore fix the whole sequence of moves she is going to make from the be-
ginning. In doing that, she will take into account the countermoves she expects the
others to do.
Because of the boundaries that constrain subjective rationality, during the game
theoretical assumptions cannot be fulfilled except in particularly simple settings. As
soon as complexity increases, the complete exploration of the game tree and perfect
foresight become too demanding assumptions for a bounded rational individual. It
seems more realistic to consider the individual exploration of the game as guided by
search procedures56 which are heuristic procedures that use cues to orient the search.
They reduce the computational complexity by introducing different parameters to
which the subjective evaluators of future configurations can be informed. A lower
amount of analysis and simpler analytical skills are involved in such a selective
exploration of the game. The players simplify the analysis by exploring only certain
segments of the game tree and by excluding others which do not seem to fulfil
their own subjective evaluators as they cannot be associated with the cues orienting
the search.
Search procedures assume the form of algorithms,57 which consist in a finite set
of instructions for accomplishing a certain task. Search algorithms prescribe ways
for choosing one out of a subset of outcomes. They are finite and work recursively.
General search procedures (“blind procedures”58 ), which enable the complete ex-
ploration of the game tree, require unlimited computational capabilities, while se-
lective search procedures (“satisficing procedures”59 ), which only insist on a part
of the game tree, heuristically orient the search according to subjective criteria and
parameters.
Puzzles, e.g. the Rubick’s cube or Sudoku, are complex problems which given
an initial configuration can generate a huge number of final configurations. The
task they posit consists in reaching a certain final state by sequentially applying the
typically very simple puzzle rules.
Similar to solving games, when solving puzzles, the individuals cannot explore
all the configurations that are possible starting from the initial state. In solving puz-
zles, the individuals inform their search to the use of different heuristics. One of the
most commonly applied is decomposition. As its name suggests, it aims to decom-
pose a complex problem into simpler sub-problems. Decomposition can be recur-
sively applied to all sub-problems, until each of them is decomposed into elementary
and very simple problems.
By decomposition local optimal solutions can at best be identified. But the dis-
covery of all local optimal solutions does not lead to finding the global optimum
56 Cf. Egidi (1992, pp. 150 ff).

57 For a discussion on definition and characterization of “algorithm” see e.g. Sipser (2006).
58 This terminology refers to Egidi (1992, p. 154).
59 Idem.
4.3 Decomposing Rationality 91
for the original problem.60 This explains the lock-in of human problem-solving that
motivates the persistence and stability of sub-optimal solutions.
A further reason for the emergency and persistence of sub-optimal solutions as
well as erroneous beliefs can be found in the path-dependency of learning. Learning
is path-dependent because the order in which alternative solutions to a problem are
considered affects the decision and influences the future problem-solving. This is
because, discovering “a solution to a problem by searching in a particular way, it
will be more likely to search in that way in future problems of the same type.”61
What emerges from the analysis of the individual approach to games and puzzles
is that human behaviour depends on the two “extreme typologies” of application of
routines and creative thought.62 The individuals facing a certain situation can learn
to implement different sequences of actions mechanically, as if they were executing
a program.63 As long as this way of behaving assures satisficing results, it is not
likely to be changed. Otherwise, the individuals will make use of their creative, ex-
plorative skills and look for new strategies, routines or methods to solve the problem
in an alternative way.
4.3 Decomposing Rationality
Kahnemann and Tversky’s seminal contribution to the analysis of the subjective

rationality64 focuses on the factors that determine the persistence and systemacy
with which violations of the rational choice paradigm occur.
Among the central findings are that thoughts differ in their accessibility (“some
come to mind much easier than others”)65 and that choice can be either based on
intuitive or on deliberate mental processes.
4.3.1 Intuition and Reasoning
Intuition and reasoning are two different modes of thinking and deciding. They can
inform judgements and decisions in that those which are taken on the basis of the
intuitive mode occur automatically and need a very short time, while those based on
reasoning, i.e. the “controlled” mode, are deliberate and slower.
60 For a demonstration of that, see e.g. Egidi (2003).

61 Cf. Cyert and March (1963, p. 174).
62 Cyert, Simon and Trow (1956) introduced in this insight the distinction between “programmed”
and “non-programmed” decisions, while in Nelson and Winter (1982) speak of “routines” and
“innovative decisions.”
63 For a discussion of elements of path-dependency see e.g Egidi and Narduzzo (1997).
64 Cf. e.g. Kahneman and Tversky (1974, 1979) and Kahneman (2002).
65 Cf. Kahneman (2002, p. 449).
PERCEPTION INTUITION REASONING

SYSTEM 1 SYSTEM 2
Fast Slow
PROCESS
Parallel Serial
Automatic Controlled
Effortless Effortful
Associative Rule-governed
Slow-learning Flexible
CONTENT
Percepts Conceptual representations

Current stimulation Past, Present and Future
Stimulus-bound Can be evoked by language
Fig. 4.3 Intuition and reasoning (Kahneman, (2002)
The distinction between “automatic” and “controlled” processes was initially de-
veloped by Schneider and Shiffrin (1977). It underlines that human thought both
involves deliberate operations and automatisms and that, in the terminology intro-
duced by Stanovich and West (2002), cognition can therefore be seen as consisting
of two systems. As shown in Fig. 4.3, the first system is responsible for intuition
(System 1), the latter for deliberate reasoning (System 2).
The conception underlying this representation of cognition is that “intuitive
judgements occupy a position [. . .] between the automatic operations of perception
and the deliberate operations of reasoning.”66
The operating characteristics of intuitive judgements are similar to those of the
perceptual processes. Both do not involve much cognitive effort and are fast and
automatic, which means that they can also be processed simultaneous to other cog-
nitive activities. Further, perception and intuition are both highly subjective, in the
sense that they cannot be easily communicated or transmitted among individuals,
and are difficult to control or modify. Intuition and perception differ in their content.
While perception is the immediate codification of information as it is originated by
a current sensorial stimulation, intuition is a cognitive mode which aims at the for-
mulation of conceptual representations that can be stored in the memory and evoked
by language.
Also reasoning aims at the elaboration of conceptual representations. Contrary
to intuition, the cognitive mode of reasoning develops on the basis of slow, serial
and controlled processes. Reasoning requires cognitive effort and therefore takes
more time. It is rule-governed and can be transmitted by learning more easily than
intuition. Further differences between perception and intuition, on the one hand,
and reasoning, on the other hand, rely in the degree to which they are voluntary and
explicit: “the perceptual system and the intuitive operations of System 1 generate
66 Cf. Kahneman (2002, p. 450).
impressions of the attributes of objects of perception and thought. These impressions

are not voluntary and need not to be verbally explicit. In contrast, judgements are
always explicit and intentional, whether or not they are overtly expressed. Thus,
System 2 is involved in all judgements, whether they originate in impressions or in
deliberate reasoning.”67
Further, System 2 supervises the mental operations of both systems and is re-
sponsible for the cognitive self-monitoring to which further application or modifi-
cation of certain behavioural patterns is informed.68 The cognitive self-monitoring
of System 2 is however quite lax. Far from being perfect, it might often allow erro-
neous intuitive judgements to enforce. The reason for that is that individuals often
tend to trust plausible judgements (as intuitive judgements inevitably are) and do
not engage in further strenuous reasoning if the reached outcome fulfils their aspi-
rations. For example,69 students of prestigious US-universities were asked to solve
the following puzzle: “A bat and a ball cost $1.10 in total. The bat costs $1 more
than the ball. How much does the ball cost?” The large majority of the students fol-
lowed the initial tendency to answer 10 cents without further checking the accuracy
of such an intuitive response.
4.3.2 Accessibility
Intuitive judgements and thoughts come spontaneously to the mind because they
are particularly “accessible” to human cognition. The key to understanding intuition
therefore lies in capturing the features that make some thoughts more accessible than
others. For this reason, the concept of “accessibility”70 which “subsumes the notions
of stimulus salience, selective attention, and response activation or priming”71 was
introduced.
For example, out of the three blocks in Fig. 4.4 three towers of equivalent height
could be erected. The three blocks differ however in their accessibility. While the
first block immediately conveys the idea of its height, which can be caught almost
intuitively, the perceptual impression suggested by the second block is related to its
area and requires more effort to distinguish its representation as a tower. Finally,
the third block’s tower has a different accessibility, which changes in dependency of
the task.
This shows that accessibility also influences the assessment of relational proper-
ties among items, e.g. making comparisons. In Fig. 4.4 the similarities between the
first and the third item seem to be more than between the first and the second.
67 Cf. Kahneman (2002, p. 451).

68 Cf. Gilbert (2002) and Stanovich and West (2002).
69 Cf. the experimental evidence in Frederick (2003).
70 Cf. Higgins (1996).
71 Cf. Kahneman (2002, p. 453).
Fig. 4.4 Shape and accessibility (Kahneman, (2002)
Accessibility is not a dichotomous concept, but develops on a continuum be-

tween “rapid, automatic and effortless operations”72 from the one extreme (as the
activities that can be ascribed to System 1 are) and “slow, serial and effortful oper-
ations”73 from the other.
Accessibility depends on the particular properties and attributes of the issue over
which a judgement has to be taken, which means that its determinants are complex
to identify and enumerate.74 Among them, physical salience, e.g. size or the position
of an item, plays a central role, since the bigger an item or the more prominent the
place it occupies, the higher its chances to attract attention are, in comparison with
a small item placed aside. The physical salience of an element positively influences
the accessibility with which the element is perceived.
Further determinants of accessibility, which can even overcome physical salience,
are motivational or emotional clues. Some items can be, independent of their physical
qualities, more accessible than others, either because they evoke emotions and
can therefore be spontaneously associated with emotional clues, or because the
individuals are motivated, for instance through instructions, habits and so on, to pay
them special attention.
Accessibility cannot be seen as an invariant property of an item and might also
change over time, since it “also reflects temporary states of priming and associative
activation.”75 Motivational and emotional clues, for example, influence accessibil-
ity much more when they are related to immediate emotions or current needs.76 At
the same time, however, accessibility is also related to “enduring operating charac-
teristics of the perceptual and cognitive system.”77
72 Cf. Kahneman (2002, p. 453).

73 Idem.
74 The determinants of accessibility are also discussed referring to Kahneman (2002, p. 453 ff).
75 Cf. Kaheman (2002, p. 454).
76 Cf. Loewenstein (1996).
77 Cf. Kahneman (2002, p. 454).
Some attributes, which have been defined “natural assessments,”78 are “routinely
and automatically registered by the perceptual system or by System 1, without in-
tention or effort.”79 Among others, they include “in addition to physical proper-
ties such as size, distance and loudness, [. . .] more abstract properties such as
similarity (e.g., Tversky & Kahnemann, 1982), causal propensity (Kahneman &
Varey, 1990; Heider, 1944; Michotte, 1963), surprisingness (Kahneman & Miller,
1986), affective valence (e.g., Bagh, 1997; Cacioppo, Priester & Bernston, 1993;
Kahneman, Ritov & Schkade, 1999; Slovic, Finucane, Peters & MacGregor, 2002;
Zajonc, 1980), and mood (Schwarz & Clore, 1983). Accessibility itself is a natural
assessment – the routine evaluation of cognitive fluency in perception and memory
(e.g., Jacob & Dallas, 1981; Johnson, Dark & Jacoby, 1985; Schwarz & Vaughn,
2002; Tversky & Kahneman, 1973).”80
The same item can make different thoughts accessible depending on the con-
text it belongs to. Obviously, the possibility of ascribing different meanings or
ambiguously interpreting such item in the rule is not perceived by the individu-
als. For example, if only one of the two rows in Fig. 4.5 would have been shown,
the ambiguity by which the second sign of each row can be interpreted, which de-
pends on the context (letters or numbers), would not have been in all likelihood
noticed.
Framing effects and prospect theory will be shortly discussed. Both can be related
to accessibility as well. Framing effects occur when the same problem suggests dif-
ferent thoughts and inspires different reactions when it is formulated in different
ways. Prospect theory relies on the observation of the different accessibility of stim-
ulations when they are expressed as absolute values rather than as changes.
Fig. 4.5 Effects of the context

on accessibility (Kahneman,
2002)
78 Cf. Kahneman and Tversky (1983).

79 Idem.
80 Cf. Kahnemann (2002, p. 454), who refers to the listing in Kahnemann and Frederick (2002).
Refer to them for the exact references.

4.3.3 Framing Effects
The phenomenon of framing regards decisions under risk and occurs when the
modification (ceteris paribus) of the context and/or the description of a problem
affects the decision. Alternative descriptions of the same problem can namely sug-
gest different impressions and therefore make different thoughts more accessible
than others. As it affects the subjective interpretation of the problem, it could
have some repercussions on the perception of outcomes and contingencies which
are associated with the different decision options. In this way, the same individ-
uals can make different choices in similar problems just because of their respec-
tive frame.
The well-known “Asian disease problem” illustrates a case in which framing
effects typically take place:81
“Imagine that the United States is preparing for the outbreak of an unusual Asian
disease, which is expected to kill 600 people. Two alternative programs to combat
the disease have been proposed. Assume that the exact scientific estimates of the
consequences of the programs are as follows:
– If Program A is adopted, 200 people will be saved
– If Program B is adopted, there is a one-third probability than 600 people will be
saved and a two-thirds probability that no people will be saved
Which of the two programs would you favour?”82
Most of the respondents confronted with this problem are prone to choose pro-
gram A, from which the clear preponderance of risk-averse individuals can be in-
ferred. This is however contradicted by the behaviour of respondents which received
the same cover story, but a different description of the two options:
– “If Program A is adopted, 400 people will die
– If Program B is adopted, there is one-third probability that nobody will die and
two-thirds probability that 600 people will die”83
In this formulation, most of the people favoured the risk-seeking alternative, i.e.
program B. This clearly violates the axiom of invariance of preferences (also known
as the extensionality axiom)84 which assumes that irrelevant features of options or
outcomes do not affect the individual preferences.
In the Asian disease problem alternative formulations of equivalent options evoke
different impressions, just because they insist either on the positive or on the nega-
tive attributes of the possible outcomes (respective survival and death). Although the
probabilities associated with the different options in the two versions of the problem
are actually the same, they are intuitively perceived in a different way. Program A is
81 Cf. Kahnemann and Tversky (1983).

82 Cf. Kahnemann (2002, p. 457).
83 Idem.
84 E.g. in Arrow (1982).
emotionally more appealing in the first version, since “the certainty of saving people
is disproportionately attractively and the certainty of deaths is disproportionately
aversive.”85
Relating framing effects with accessibility and recalling the considerations on the
imperfection of control by System 2 over System 1’s operations, it is not surprising
that the frame of a problem can even bias the choice of expert decision-makers.
McNeill, Pauker, Sox, and Tversky (1982) show for example that if asked to choose
between surgery or radiation therapy, both professionals’ and patients’ decisions
could be framed by describing outcomes in terms of survival or mortality rates.
Framing effects can bias problem-solving activities, as well. For example, it has
been observed86 that the representation of a problem affects skill transferring. The
amount of skill transfer among problems depends more on the perceived rather than
on the objective similarities between them, so that isomorphic problems may be
perceived as different and unrelated to each other. There is evidence proving that
individuals are susceptible to the frame in which a problem is presented as they tend
to adapt passively to its proposed formulation and do not reduce the problem to its
stylized, canonical representation which would reveal eventual analogies with other
(already known or solved) tasks.
The framing effects represent a systematic violation of the rational choice theory
embodied in its invariance assumption, which means they seriously challenge the
descriptive validity of the neoclassical approach and give a special emphasis to the
role of accessibility in human problem-solving and decision-making.
4.3.4 Prospect Theory
A further central finding which questions the validity of rational choice theory is the
reference-dependency of perception, on which basis the framework of the prospect
theory relies.
Perception is reference-dependent, because “the perceived attributes of a focal
stimulus reflect the contrast between that stimulus and a context of prior and con-
current stimuli.”87
This property of perception can be illustrated considering the two enclosed
squares of Fig. 4.6 whose brightness is erroneously perceived as differing, because
of the brightness of the surrounding area. This influences perception in that it fixes
a sort of benchmark to which perception refers, which is called “reference value” or
“adaptation level.” Not only current stimuli reaching the brain determine perception,
but also prior stimulations as well as the history of adaptation.
Being guided by perception, the evaluation of decision-outcomes is reference
dependent, too. This contradicts what is posited by the expected utility theory.88
85 Cf. Kahnemann (2003, p. 457).

86 For an overview, see e.g. Mark and Eysenck (2005).
87 Cf. Kahneman (2003, p. 459).
88 Expected utility theory has been first formulated by Bernoulli (1738).
Fig. 4.6 Reference-dependency of perception (Kahneman, 2002)
This approach, which still remains the dominant approach to risky choice, assumes
a logarithmic form of the utility function of wealth, postulating therefore that an
increase in utility due to an increase in wealth is indirectly proportional to the initial
wealth state.
In this framework, the maximization of the utility of wealth implies risk aversion
and yields an explanation for the differences in the risk attitudes of individuals with
differing initial states of wealth (high or low, i.e. rich or poor). Although Bernoulli’s
expected utility theory considers that different states of wealth have an effect on
the individual attitudes toward risk, it does not relate risk attitudes to the reference
provided by the initial state of wealth and does not account for the different way
individuals are prone to experience losses instead of gains.
The centrality of the reference value in determining the individual risk attitude is
proved by empirical and experimental evidence,89 which reveals how “the carriers
of utility are likely to be gains and losses rather than states of wealth.”90 Such evi-
dence can be related to the principle of reference-dependency of perception, whereas
“the effective stimulus is not the new level of stimulation, but the difference between
the existing adaptation level.”91 Modelling risky choice in a psychological realistic
way means to allow for the reference-dependency of perception.
In this spirit, Kahneman and Tversky (1979) developed the framework of the
“prospect theory.” Prospect theory relies on the idea “that carriers of utility are
changes of wealth rather than asset positions”92 which implies among other things
“that choices are always made by considering gains and losses rather than final
states.”93 The different accessibility of gains and losses frames the perception of
changes in wealth, as it is intuitively illustrated by the two problems discussed
in Fig. 4.7.
89 See e.g. Kahneman and Tversky (2002).
90 Cf. Kahneman and Tversky (2002, p. 461).
91 Cf. Kahneman and Tversky (2002, pp. 460, 461).
93 Idem.
Problem A Problem B
Would you accept this gamble? Which would you choose?
50 % chance to win 150 $ Lose 100 $ with certainty

50 % chance to loose 100 $ OR
50 % chance to win 50 $
50 % chance to loose 200 $
Would your choice change if your overall wealth Would your choice change if your overall wealth
were lower by 100 $? were higher by 100 $?
Fig. 4.7 The wealth frame (Kahneman & Tversky, 2000)
value
losses gain
Fig. 4.8 The value function (Kahneman & Tversky, 2000)
It can be intuitively inferred that most of the individuals who face problem A will
in all likelihood not accept the gamble and therefore reveal risk aversion.
The lottery presented by Problem B looks appealing. Evidence confirms that the
majority of the individuals confronted with a similar choice typically accept the
gamble, thus behaving in risk-seeking way. The evidence further shows that even a
$100 increase in wealth does not induce them to revise their choice.
The expected utility theory considers problem A and B to be essentially equiva-
lent and cannot therefore account for the differing risk attitudes revealed by the in-
dividuals in these two settings. Instead of that, Kahneman and Tversky’s approach,
modelling “an alternative theory of risk, in which the carriers of utility are gains
and losses – changes of wealth rather than states of wealth,”94 provides a suitable
framework for explaining how the perception of gains and losses frame the individ-
ual risk propensity.
The prospect theory predicts that the individuals favour risk averse choices when
gains are involved, while it expects them to prefer risk-seeking behaviours when
losses are involved. The value function represented in Fig. 4.8 visualises these re-
sults.
The value function is defined on gains and losses. While it is concave in the
domain of gains, expressing therewith risk aversion, it is convex in the domain of
losses. The value function models the different attitudes toward risk in a dichoto-
mous way, in that it is sharply kinked at the origin, which constitutes the reference
point for choice. The slope of the curve, which is steeper for the losses than for the
gains, further indicates a higher sensitivity of risk propensity in the domain of losses
rather than in that of gains.

Chapter 5
Heuristics, Biases and Methods for Debiasing
Prediction is involved in many recurrent tasks individuals are confronted with and
can be seen as the result of the interaction between “judgement, intuition, and edu-
cated guesswork.”1 Even when forecasts rely on mathematical methods the central
role of intuition cannot be denied as it supervises e.g. the choice of variables that
belong to the model, their initial value and their functional specification.
This chapter deepens a central aspect for human cognition and problem-solving,
namely that individuals make use of bounded rational heuristics for taking decisions
under uncertainty. Heuristics are simplified procedures for assessing probabilities.
They are based on rules of thumb. They rely on mental clues which selectively
orient the search process and enable the individual to reach her goals when time,
informational and computational capabilities are constrained.
Although in some cases bounded rational heuristics can be made responsible for
the sub-optimality of outcomes and for behavioural biases. In some other cases it
represents an essential support for carrying on inference when complexity overloads
the individual cognitive and computational capabilities. It enables the individual to
reach better solutions than otherwise.
There are mainly two different approaches to subjective judgement and bounded
rational heuristics, namely the “heuristics and biases” approach, pioneered by Kah-
neman and Tversky,2 and the “ecological rationality” approach, with Gigerenzer3 as
one of its most influential proponents.
While the “heuristics and biases” approach underlines “that inference is system-
atically biased and error-prone, powered by quick and dirty cognitive heuristics,”4
the “ecological rationality” approach rather insists on ecological aspects of fast and

2 See e.g Kahneman (2002), Kahneman and Tversky (1974, 1979), (2000) and Kahneman, Slovic,
and Tversky (1982).
3 See e.g. Gigerenzer, Todd, and the ABC Research Group (1999).
4 Cf. Chase, Hertwig, and Gigerenzer (1998, p. 206), referring to Kahneman, Slovic, and Tversky
(1982).
102 5 Heuristics, Biases and Methods for Debiasing
frugal heuristics, which exploit the regularities of the physical and social environ-
ment and enable individuals to make inferences which rely on a manageable amount
of clues.5
The essential difference between the two approaches consists in the interpreta-
tion of biases as systematically affecting the individual judgement or rather related
to exceptions and therefore not seriously compromising the human capability of
assessing inferences.
Heuristics are a constitutive feature of human thought as they are deeply inter-
nalised and constitute particularly resistant feature of the human cognition. They
cannot be easily modified, eliminated or substituted, even when they lead to erro-
neous judgements which inform sub-optimal choices. This aspect emerges signifi-
cantly from the attempts at debiasing the subjective judgement and behaviour, which
often reveals a moderate to high resistance of several perceptional and judgemental
biases.
As follows, the main features of bounded rational heuristics will be underlined.
The heuristics of representativity, availability, anchoring and adjusting will be deep-
ened. The analysis will focus on some of the aspects underlined by the heuristics
and biases approach but also consider some of the critiques moved by the ecologi-
cal rationality approach. An overview on the research on debiasing, whose methods
and main results will in particular be related to the recursivity of economic theories
and the possibility of testing their absorption, will conclude the chapter.
5.1 Bounded Rational Heuristics
Many decisions have to be taken under conditions of uncertainty, i.e. where different
events are likely to occur. In similar cases decision-making will be informed to the
estimation of the likelihood with which different outcomes can be predicted.
The bounded rational revolution deeply questions the conception of human in-
ference as adherent to the laws of logic and probability theory, as it states the cen-
trality of bounded rational heuristic rules in guiding the individual assessment of
probabilities.
The Greek word “heuristic” refers to something which serves to find out or dis-
cover.6 Its appearance into English can be traced back to the early 1800s where
it was applied to depict useful processes alternative to the postulates of logics
and probability theory for solving puzzling problems. The concept of heuristic re-
ceived a negative connotation in Einstein’s seminal paper of 1905 in which the term
“heuristic” was used for indicating incomplete and erroneous ideas. It was never-
theless interpreted as useful and supportive for directing human thought. Simon and
Newell kept on with the Dunckerian interpretation of heuristics as (mostly useful)
5 Cf. Chase, Hertwig, and Gigerenzer (1998).

6 This introduction on the notion of “heuristic” refers mainly to Gigerenzer and Todd (1999,
p. 25–29).
5.1 Bounded Rational Heuristics 103
tools for orienting the search for information7 and modelled heuristics as compu-
tational methods.8 In this way they underlined the supportive role heuristics may
assume for human cognition and decision-making as well as the character of rules
of thumb, simplified procedures and useful mental devices. In the 1970s, however,
a more negative connotation of this concept found its re-emergence. It can be as-
cribed to its usage in studies analysing the psychology of decision-making, which
adopted the term heuristic to depict mostly dispensable cognitive processes peo-
ple erroneously choose to apply instead of relying on logic and probability the-
ory.9 However, the almost synonymous association between heuristics and biases
was not in the original spirit of the homonymic research program of Kahneman and
Tversky (1974). Their research was rather conformed to the consideration that even
if heuristics may systematically bias the human cognitive and decisional processes,
they nevertheless represent essential tools for supporting decision-making in situ-
ations in which the boundaries posited to the subjective rationality do not allow a
decision to be taken on a different basis.
Profoundly different spirits respectively animate the “heuristic and biases” and
the “fast and frugal heuristics” approaches: whilst the latter “sees heuristics as the
way the human mind can take advantage of the structure of information in the en-
vironment to arrive at reasonable decisions,”10 the first insists more on those cases
in which heuristics can be viewed rather “as unreliable aids that the limited hu-
man mind too commonly relies upon despite their inferior decision-making per-
formance.”11 Despite that, both programs stress the crucial role heuristics plays in
guiding human thought and cognition.
5.1.1 Building Blocks of Bounded Rational Heuristics
Heuristics supports bounded rational decision-making in all its different phases.

It specifies particular different principles for guiding and stopping search, whose
results will then inform the decision. The principles posited by bounded rational
heuristics tend to orient the selective search for information on the basis of cues
rather than relying on extensive computational procedures. Searching for cues can
be either random or ordered; whereas for an ordered search some simplified heuris-
tic criteria are applied.
The search has to be selective oriented and stopped because of internal (cog-
nitive and computational) and external (referred to time and costs of search) con-
straints. Heuristic principles establish some criteria for terminating the search at
7 Gigerenzer and Todd (1999) refer in this insight to Duncker (1935).

8 See e.g. Simon (1955).
9 Gigerenzer and Todd (1999) remands for the early usage of the word “heuristic” to Groener
(1983), Polya (1954) and for its 1970s usage to Kahneman and Tversky (1974).
11 Idem.
an appropriate point. These principles can either be cue-based12 or referred to an

individually specified aspiration level.13 They provide the decisive advantage of
dispensing the individuals of evaluating any cost-benefit trade-off, to which opti-
mization under constraints would be on the contrary based.
After information has been (selectively) collected and the search for it stopped,
a final set of heuristic principles orients the evaluation of the results of the search
process and informs the inference assessment and the individual choice.
These different sets of principles are separable and can be combined with each
other. They can be considered as quite autonomous blocks, building the toolbox of
the subjective rationality. This feature represents a further simplification which is
introduced by bounded rational heuristic reasoning, which results in the possibility
of reducing to an extremely manageable amount the numbers of sets of heuristic
principles that has to be stored in the individual memory and applied for making
decisions under risk. The possibility of a combination among different sets of prin-
ciples additionally enhances the adaptability of heuristics to a broader range of
situations and problems.
Heuristics can in other words be combined, nested and integrated. This allows
for the reduction of the capabilities involved in bounded rational decision-making
and problem-solving.
5.1.2 Main Features of Bounded Rational Heuristics
Heuristics is a product of human cognition. Shedding some light on the mechanism

through which it emerges and by which it is stored in the cognitive repertoire would
involve deepening of the processes which rules human cognition and its extremely
complex interaction.
Heuristics can however be simplified to be stored in the cognitive repertoire by
means of choosing a certain option (it won’t be discussed whether this happens ran-
domly or following pre-existing mental clues here) and then observing its outcome.
Heuristics can thus be assumed to be learned by experience14 and specifically
by a trial-and-error process. The results of which originally refer to the specific
problem faced and are then inductively generalised to be applied to a range of other
situations that are perceived to be similar. The inductive nature of the process by
means of which heuristics is inferred from experience and extended to a certain set
of situations implies that heuristics has the properties of context-specificity, task-
generality and robustness.
Bounded rational heuristics are specific in that they are highly context dependent.
Because the individual understanding of a problem is ruled by its accessibility, its
objective features are not the only factors that could affect the way it is perceived.
12 Cf. e.g. Gigerenzer and Goldstein (1996).

13 Cf. e.g. Simon (1956a, 1956b, 1990) and Todd and Miller (1999).
14 Cf. Campbell (1969).
5.1 Bounded Rational Heuristics 105
Exterior elements might play a role, too, because the way a problem is displayed
and presented can frame its interpretation among individuals.15
The property of context-dependency does not rule out the generality of heuristics
over tasks; in absence of that there should be as many rules as concrete situations.
A certain heuristic can therefore be said to be specific but general at the same time,
because given the set of the problem’s isomorphs16 the heuristic will be applied to
the subset of isomorphs that are actually perceived as such. To all the other iso-
morphs different heuristics will be applied, thus yielding different solutions and
outcomes.
The property of generality is articulated over different degrees of dependency of
the number of tasks that are solved by means of applying the same rule. Generality
can be rated on a continuum where the one extreme is represented by heuristics
with a very restrictive domain of application, up to heuristics defining rules on how
to generate rules.
The heuristics of representativeness, availability, anchoring and adjusting can be
situated at this second extreme, as they “direct the way in which specific rules can be
formed to solve problems.”17 They are therefore also labelled as “meta-heuristics.”
Meta-heuristics allows for the high level of generality which is implied by the
task of rule generation and at the same time “for the important effects of context,
wording, response mode, and so on.”18
The particular feature of bounded rational heuristics which makes it suitable for
leading the individual decision-making in an ecological way, i.e. in a way that fits the
structure of the environment, is constituted by its solution of the trade-off between
speed and accuracy of judgement in which any deductive reasoning is inevitably
trapped. On this basis heuristics can often be at the same time fast and accurate to an
adequate degree. This is because heuristics is articulated on a different dimension:
the trade-off between specificity and generality.
As a consequence, heuristics reveals different degrees of robustness or strength,
which diminishes as its fitting to the structure of the environment increases. In par-
ticular, the strength of certain heuristic influences its reinforcement. This is ruled by
the feedback the heuristic received as a result of the individual experience, which
can vary according to how and how often such a feedback occurs, which size it has,
with which magnitude it is perceived and which methods are implemented by the
individuals for checking on the outcomes achieved by means of rule application.
A critical point in this insight is that even inadequate and incorrect heuristics
can be reinforced. This can occur when the individuals are unaware of the task
structure and cannot thereforeperceive alternative or more appropriate, better ways
15 Experimental evidence on that is conspicuous and wide accepted, see e.g. the contributions of
Grether and Plott (1979), Lichtenstein and Slovic (1971), Simon and Hayes (1976), Tversky and
Kahneman (1980).
16 The expression “problem isomorphs” was introduced by Simon and Hayes (1976).
17 Cf. Einhorn (1982, p. 271).
18 Cf. Einhorn (1982, p. 271).
of solving the task. In a similar setting the concrete outcomes cannot be related to
any superior benchmark of a solution, but can be only adjusted toward the individual
aspiration level.
After plotting the experienced action-outcome combinations on an imaginary di-
agram, the outcomes will be compared to the individual aspiration level and con-
sequently assessed either to the category of success or failure. Such an assessment
can however be biased for several reasons: because of the absence of a complete
map of experienced action-outcome combinations, because such an incomplete map
becomes in the rule only sequentially available and is sometimes widely scattered
over time, as well as because the systematic search for evidence disconfirming the
heuristic applied is typically missing or is at least very rare.
Consider for example a waiter serving in a crowded restaurant.19 Given the im-
possibility of offering good service to all customers, he might decide to dedicate
more attention to those customers he assesses to be more likely to leave a higher
tip. Treatment effects however (i.e. good service influences the tip other things be-
ing constant) elevate the probability of the waiter to confirm himself in his heuristic
following. It does not seem plausible to assume from the waiter that he disentangle
the treatment effects and give good service to some of those customers he does not
expect to pay a generous tip. This would require awareness of the task structure on
the one hand and is on the other hand not likely to be practised because of lacking
motivation: why should the waiter risk a loss of income for doubting a rule that
seems to work?
From the properties of specificity and generality of bounded rational heuristics as
well as from the modalities by means of which outcomes are subjectively assessed, it
can be inferred that the validity of heuristics should be evaluated according to corre-
spondence rather than to coherence criteria. This means that its evaluation should in-
sist on “measures that relate decision-making strategies to the external world rather
than to internal consistency, such as accuracy, frugality, and speed [. . .].”20 Heuris-
tics has the function of enabling reasonable adaptive inferences given the concrete
structure of the environment, within time and informational constraints, so that as
long as it fulfils this function there is no reason to require internal coherency.
The purpose of this chapter is to show evidence of some parallels between the
analysis of theory absorption and that of debiasing. Therefore, the discussion of
bounded rational heuristics which follows mostly relies on the heuristics and biases
approach and presents only a selection of cases of heuristics which lead to biased
behaviour.
For a discussion on the differences and similarities between the heuristics and
biases approach and the ecological rationality approach, attention should be directed
to the interesting debate between Kahneman, Tversky and Gigerenzer.21
19 Cf. Einhorn (1982, p. 282).

21 For a further critical overview, see Gigerenzer and Todd (1999, p. 28).
5.2 Heuristics and Biases 107
5.2 Heuristics and Biases
According to the classification proposed by Kahneman, Slovic, and Tversky (1982),

representativeness, availability and anchoring heuristics will be presented as the
basilar principles supporting bounded rational probability assessment and value pre-
dicting. The most frequent biases that can be associated with the application of
heuristics will be mentioned as well.22
5.2.1 Representativeness Heuristic
Common tasks involving likelihood estimations can be of the following type: “What
is the probability that A belongs to B?”; “What is the probability that A originates
from B?”; “What is the probability for B to generate A?”23
In order to solve similar problems, the representativeness heuristic is typically
applied. This is a heuristic which evaluates probabilities by the degree to which A is
representative of B. On this basis, whenever A’s perceived affinity or similarity with
B is high (respective low), the probability for B to generate A is judged to be high
(respective low) as well.
Choosing the affinity between events as a criterion for assessing probabilities can
lead to systematic mistakes. This occurs in particular when providing the individu-
als with the description of a certain phenomenon where no specific evidence is given
and asking them to assess the probability of a list of events concerning the phenom-
enon described. They will tend to base their inference on the representativeness of
the different events with the description provided. This occurs for example in the
many cases in which people make judgements on the basis of cliché, rather than on
real informative elements: take the example of the higher propensity to associate a
researcher or a scientist with the cliché of an introverted and reflexive person, rather
than with a sociable funny and easy-going guy.
Representativeness heuristics has been explained as the application of “inten-
sional” instead of “extensional thinking”24 This happens when problems are repre-
sented by means of referring them to individual mental models, to prototypic and
similar representations, instead of being conceptualized according to formal notions
and parameters.
When representativeness is relied on, other factors which are significant for in-
ferring probabilities are neglected. In this way, it can be shown that individuals are
systematically prone to the insensitivity to prior probabilityof outcomes (base-rate
22 This discussion of heuristics and biases mostly relies on the analysis of Tversky and Kahnemann
(1982, p. 3 ff).
23 Cf. Tversky and Kahnemann (1982, p. 4).
24 Cf. Kahneman and Tversky (1983).
bias), to sample size (law of small numbers) and to predictability. In addition indi-
viduals tend to mis-conceptualize chance and regression and underline the illusion
of validity.25
5.2.2 Availability Heuristic
In estimating probabilities, a typical clue which individuals rely on is availability.

Availability implies that the likelihood of certain events is estimated by recalling the
frequency of their occurrence among their own acquaintances.
Even if, on the one hand, availability represents a useful criterion for inferring
probabilities (since “instances of large classes are usually reached better and faster
than instances of less frequent classes”),26 on the other hand, it can be affected
by factors that do not depend on probability and frequency. The use of availabil-
ity heuristics can therefore systematically lead to biased judgements which can be
e.g. alleged to the retrievability of instances, to the effectiveness of a search set, to
imaginability as well as to illusory correlation.27
5.2.3 Adjustment and Anchoring
Anchoring refers to the tendency of overly relying on a specific piece of information

and consequently to adjust the intuitive judgement to the value provided by that
piece of information which then serves as a basis for inferring other elements of the
problem.
The adjustment and anchoring heuristic implies that the individuals anchor their
judgemnt to an implicitly suggested reference point for then adjusting toward that
point. The tendency to anchor judgement persists even when no reference point is
given because individuals are prone to estimate a reference value. They might even
infer this value on the basis of incomplete computations.
Among the biases that can be related to the adjustment and anchoring heuristic,
insufficient adjustment can be mentioned, as well as biased evaluation of conjunctive
and disjunctive events, and erronous assessment of the distribution of subjective
probability.28
25 All of those biases are threaten in more details by Tversky and Kahnemann (1982). The same
applies for the biases mentioned in Sects. 2.2 and 2.3 of Chap. 5.
27 See for more details Tversky and Kahnemann (1982).
28 See note 356.
5.3 Debiasing 109
5.3 Debiasing
Included within debiasing methods are all the strategies for testing the robustness of
an observable bias by attempting to eliminate it under controlled conditions.
The robustness of a bias can be tested by means of controlled modification of the
design in which such bias is observable. The sense of such a controlled modification
of the design is to push it towards and beyond the limits of its validity, i.e. up to its
specimen’s failure. Borrowing a typical engineering expression, debiasing could be
interpreted as a sort of destructive test which reveals under which conditions the
design fails or at what point unbiased judgement and decision-making emerge.29
The search for the conditions under which the biased behaviour disappears (or is
reduced in occurrence) allows for the mechanisms to be isolated which are respon-
sible for the bias.
It is in this sense that research on debiasing constitutes one of the possible ways
to investigate the mechanisms ruling human perception, cognition and judgement.
Debiasing methods can further represent useful devices for testing the hypothesis
concerning such cognitive mechanisms and mental structures. In this insight, both
successes and failures of debiasing may yield relevant findings, in that “(a) failure
helps clarify the virulence of a problem and the need for corrective or protective
measures, and (b) the overall pattern of studies is the key to discovering the psy-
chological dimensions that are important in characterizing real-life situations and
anticipating the extend of biased performance in them.”30
A first affinity between the research on debiasing and the investigation of the self-
referential effects of a theory can be stated in this insight: a method for debiasing
can only be effective if it is able to capture the psychology underlying the individual
choice; a theory won’t have self-referential effects and cannot be absorbed among
certain individuals if it is not able to penetrate the way such individuals interpret the
economic problem they face. This will be discussed in more details in chapter 6.
The different debiasing methods can be classified “according to their implicit
allegation of culpability,”31 as biases can be induced by the formulation or the ac-
cessibility of a task. They can also be due to erroneous judgements. However biases
can also originate from the mismatch between task structure and judgement. Cor-
responding to the factors the bias is supposed to be the result of, debiasing can be
strived for by means of applying different strategies as summarized in Table 5.1.
The specific structure of a task could be responsible for biased behaviours if it is
e.g. unfairly formulated, based on confusing instructions, unfamiliar stimuli, lacking
incentives to perform well, disbelief in experimental assertions on the task nature,
mistrust of the declared payoff-structure and so on. Whenever similar conditions
occur it is the case of an unfair task. Checking on the conditions which are per-
turbing the behaviour of the individuals confronted with an unfair task would serve
29 Cf. Fischoff (1982b, p. 422).

30 Cf. Fischoff (1982b, p. 423).
31 Cf. Fischoff (1982b, pp. 422, 423).
Table 5.1 Debiasing methods according to underlying assumption (Fischoff, 1982b)

Assumption Strategies
Unfair tasks Raise stakes
Clarify instructions/stimuli
Discourage second-guessing
Use better response modes
Ask fewer questions
Misunderstood tasks Demonstrate alternative goal
Demonstrate semantic disagreement
Demonstrate impossibility of task
Demonstrate overlooked distinction
Faulty judges
Perfectible individuals Warn of problem
Describe problem
Provide personalized feedback
Train extensively
Incorrigible individuals Replace them
Recalibrate their responses
Plan on error
Mismatch between judges and task
Restructuring Make knowledge explicit
Search for discrepant information
Decompose problem
Consider alternative situations
Offer alternative formulations
Education Rely on substantive experts
Educate from childhood
to improve the scientific hygiene of the study, rather than attempting a debiasing
operation in a narrow sense.
It can however be difficult to draw a clear dividing line between a misleading de-
scription which artificially biases the individual judgement and a description which
frames the individual interpretation of the situation faced. In this sense, biases can
also be due to the specific description or formulation of the task, which is susceptible
to misunderstanding by the individuals. Task misunderstanding can be for example
checked and eventually avoided by using alternative descriptions, underlying differ-
ent key words, pointing explicitly at alternative goals, or by advising the individuals
to concentrate on overlooked distinctions.
“Once the task has been polished and the bias remains”32 a second approach
to debiasing can be followed. The responsibility is then definitively alleged to the
individuals performing the task.
The individual judgement can be either perfectible or incorrigible. Whenever the
latter case applies, the subjective judgement is not sensitive to any debiasing training
and device. The shortcomings introduced by the bias can only be overwhelmed by
flanking the decision-makers with artificial support.
32 Cf. Fischoff (1982b, p. 426).
5.3 Debiasing 111
Obviously, in case of incorrigible judgement, testing for debiasing strategies

would not make much sense for deepening the analysis of the individual bounded
rational judgement. Debiasing attempts make much more sense in the case of per-
fectible judgements. In these cases, debiasing can be achieved by providing the indi-
viduals with artificial experience, whereas searching for the most suitable methods
(i.e. that can be understood, internalised and applied or in other words “absorbed”
among the individuals) enables some light to be shed on the mechanisms ruling
perception and cognition. Experience can be simulated and artificially provided to
the individuals for example through different training programs. These training pro-
grams consist of warnings about the possible bias occurrence and offer personalized
feedback or problem descriptions.
As it will be discussed later in more details (see chapter 7), the absorption of a
theory can be empirically tested by means of providing the individuals with theoret-
ical information concerning the concrete situation they face. The analysis of theory
absorption can therefore be methodologically related to the debiasing of perfectible
judgements, in that it similarly implies the training of the individuals in order to
support their subjective rationality.
A third source of biases is constituted by the mismatch between judgement and
task. The two possible approaches to debiasing that can basically be followed when
a bias is implied by the discrepancy between judgement and task are “restructuring”
and “education.”
The procedure of restructuring can be endeavoured if the compatibility between
individual skills and structure of the task is potentially given, but fails for some
reasons to get exploited. If there is a subset of individuals who are able to exploit
the compatibility between task and skills and another group who are not capable,
debiasing can be achieved by educating the unsuccessful respondents.
Restructuring can in other words be applied if the respondents are assumed to
be in possession of the skills required to solve the task, but they are observed so
that not make use of them in solving the task. Inviting the respondents to express
their knowledge and/or reasoning explicit, decomposing the problem, considering
alternative situations or formulations of the task, making discrepant information
more accessible etc. are among the different restructuring strategies that can be
mentioned.
Educating the respondents is a debiasing procedure that suits the cases in which
a bias is assumed to occur even if the task does not involve cognitive and computa-
tional capabilities that are available to the respondents. Such an assumption can be
e.g. tested by comparing the biased performance of the respondents with the bench-
mark of individuals that are already acquainted with the task and have therefore
developed a certain expertise. If this assumption seems to apply, debiasing could be
achieved by means of “educating” the respondents. “Education” is meant here to
differ from training measures concerning the faulty judgement category, in that it
focuses “on developing general capabilities rather than specific skills.”33
A similar educating procedure can be specified for the analysis of the self-
referential effects of a theory and its absorbability, as well. As it will be specified in
33 Cf. Fischoff (1982b, p. 427).
the coming chapters (Chaps. 6 and 7) testing for theory absorption will be based on
the procedure of providing the respondents with theoretical information concerning
the concrete task to be solved.
Although some general patterns or approaches to debiasing can be inferred, con-
crete debiasing procedures essentially differ from one bias to the other because of
specificity and task dependency of heuristics. For this reason, as well as for the dif-
ferent attention which has been paid in the literature to the diverse cognitive biases,
it is not easy to organize a systematic overview on debiasing. In addition, a certain
degree of arbitrariness is always present when it comes to evaluating the relevancy
of a certain debiasing attempts. Whether or not the efficacy of a debiasing experi-
ence or its methodological fundaments should be privileged, can be argued.
Three considerations essentially oriented the selection of the studies that are go-
ing to be mentioned, namely (1) to systematically privilege, even when it is at the
expense of comprehensiveness; (2) to focus on the contributes providing some ele-
ments that can be interpreted as preliminary for the analysis of theory absorption;
(3) in respect of the criteria posited by Fischoff (1982a, 1982b), to restrict the atten-
tion to studies published after peer review which offer some empirical evidence so
that both theoretical suggestions and anecdotic evidence are ruled out.
A selection of debiasing studies which meet these requirements and focus re-
spectively on the representativeness, availability and adjustment meta-heuristics will
follow.
5.3.1 Debiasing the Representativeness Heuristics
When debiasing the representativeness heuristic, two alternative approaches are

mainly followed: the procedures either aim at affecting the representation of in-
formation or at the training of the respondents.
5.3.1.1 Debiasing the Insensitivity to Prior Probability of Outcomes
Focussing on the insensitivity to the prior probability of outcomes, Rakow, Harvey,

and Finer (2003) demonstrated for example that it is possible to reduce the occur-
rence of the base-rate bias if the respondents are allowed to discuss the situation in
small groups. The study proved group discussion to draw the respondents’ attention
to the base-rate information provided because it stimulates the individuals to think
more carefully about the information they received. Moreover, a second experiment
revealed the beneficial effects of providing the respondents with the range of proba-
bility responses for the task. Of the range of probability responses, graphical instead
of numerical representation yielded for different results in improving the accuracy
of judgement.34
34 Cf. Rakow, Harvey, and Finer (2003).

5.3 Debiasing 113
According to Gigerenzer, Hell, and Blank (1988) the systematic neglection of

base-rate information can be alleged to the internal problem of representation rather
than the use of the representativeness heuristic. This implies that “by manipulating
presentation and content, one can elicit either base rate neglect or base rate use,
as well as points in between”35 and reveals therefore that “representativeness is
neither an all-purpose mental strategy nor even a tendency, but rather a function of
the content and the presentation of crucial information.”36
Roy and Lerch (1996) rely on three alternative strategies for correcting the insen-
sitivity to the prior probability of outcomes: different presentations of the informa-
tion, training of the respondents in strategic devices for processing information and
replacement of human decision-makers with a model or system following normative
rules of probability theory.
Moreover, while Cole (1989) could diminish the insensitivity to priors by edu-
cating the respondents through the provision of feedback information, Gebotys and
Claxton-Oldfield (1989) lowered the impact of base-rate neglecting by adding to the
theoretical knowledge of the respondents. By a follow-up test 5 weeks after the ex-
periment they could additionally prove that the training group maintained a superior
performance.
5.3.1.2 Debiasing the Conjunction Fallacy
The conjunction fallacy occurs when specific conditions are assumed to be more
likely than general ones because they are intuitively judged to be more representative
than the general ones.
The attempts at debiasing the conjunction fallacy are mostly centred on pointing
at the concept of “specificity” as a subset of that of “generality,”37 as well as on
increasing the respondents’ acquaintance with elementary notions of set theory.
For example, Agnoli and Krantz (1989) developed a training method which com-
bines external pictorial representations (Venn diagrams) with learning by doing. As
their results suggested, the efficacy of debiasing seems to be strived by inducing the
individuals to shift from the mode of intensional to that of extensional thinking. In
the end this can be related to a reduction in the intuitive components involved in the
judgement.
In 2005 a pilot experiment was conducted on undergraduate students at the Uni-
versity of Dresden which aimed at debiasing the conjunction-effect fallacy by means
of providing the respondents with theoretical information. The experiment merged
typical elements of debiasing (in form of graphical aids) with abstract theoretical in-
formation and focussed on the analysis of theory absorption. As a result, the almost
complete elimination of the conjunction-effect bias could be achieved. Further de-
tails on this pilot experiment will be discussed shortly in Sect. 4.1 of Chap. 7.
35 Cf. Gigerenzer, Hell, and Blank (1988).

36 Idem.
37 Cf. e.g. Moutier and Houdé (2003).
5.3.2 Debiasing the Availability Heuristic
Stapel, Reicher, and Spears (1995) codified and tested different moderators on
the impact of biases connected with the availability heuristic. In doing that they
emphasized the extreme context dependency of heuristics and judgement and in-
sisted on the perceived relevance of the information provided for the task to be
solved as well as on its different memorability. What they showed was that simply
introducing a different cover story for the same task can lead to different judge-
ments, as contextual factors can steer the use of certain information. They further
provided evidence for high context-dependency of the perceived relevance of infor-
mation, which rules its memorization and therefore influences its availability.
The hindsight bias expresses the propensity to see past events as predictable and
reasonable as well as the overconfidence in the individual’s own predictive abilities.
The hindsight bias can be ascribed to a plurality of psychological factors (e.g. focal
thoughts, accessibility, selectivity of memory, overconfidence etc.) and is based on
complex psychological mechanisms that are deeply rooted in the human mind. For
this reason, the insight bias occurs highly systematically and is particularly resistant
to debiasing.
Debiasing procedures developed on the content of the prediction whose percep-
tion is biased by hindsight, reveal only a moderate efficacy,38 while procedures
which act on the interaction between focal thoughts and accessibility39 enhance
more effecting to lower the bias’ occurrence.
5.3.3 Debiasing Adjustment and Anchoring
Mumma and Wilson (1995) developed and tested three procedures for debiasing
anchoring effects in clinical-like judgements, each of which correspond to an alter-
native explanation of anchoring.
The first procedure (the so called “bias inoculation approach”) consists in training
the respondents about the anchoring heuristic and providing them with feed-back
information on the primacy bias.
The second procedure (the “encoding elaboration model”) invokes the role
played by opposite information in inducing the respondents to realize differences
among initial and late information. This second procedure has also been applied
(in form of the so-called “consider the opposite” procedure) to the debiasing of the
overconfidence and hindsight bias.40
The third procedure tested by Mumma and Wilson (1995) finally focuses on the
“attention decrement” as a possible reason for the anchoring effect and accordingly
consists of asking the respondents to write some notes on the information received.
38 See e.g. Fischoff (1982a) and Pohl and Hell (1996).
39 Cf. e.g. the contribution of Sanna and Schwarz (2004).
40 See e.g. Russo and Schoemacher (1992) for the overconfidence bias and Sanna, Stocker, and
Schwarz (2002) for the hindsight bias.

5.4 Concluding Remarks on Debiasing and Some Implications for Theory Absorption 115
When applied to single-cue anchoring, all these three procedures showed dif-
ferentiated but altogether encouraging debiasing results if applied to single-cue an-
choring. They revealed almost no efficacy if applied to the more complex case of
sequence-anchoring.
5.4 Concluding Remarks on Debiasing and Some Implications

for Theory Absorption
The studies on debiasing that have been sketched so far generally reveal a moderate
robustness of the cognitive biases. Overall, the reviewed contributions provide fur-
ther evidence for the systematicity with which biases occur. In addition they show
how, in most of the cases, biases cannot be reduced to a mere artefact of a faulty task
or interpreted as a result of the triviality of the task which causes a lack of motivation
and interest among the respondents. Debiasing efforts could be typically proven to
achieve significant modifications in the biases’ occurrence whenever they were able
to modify the subjective approach to the task by acting on the mechanisms ruling
the individual perception of the task.41
Whereas the research on debiasing fixes some common patterns for elaborating
corrective procedures, it still leaves several questions open. The overall result which
emerges is that people “are skilled enough to get through life, unskilled enough
to make predictable and consequential mistakes; they are clever enough to devise
broadly and easily applicable heuristics that often serve them in good stead, unso-
phisticated enough not to realize the limits of those heuristics.”42
Whereas given the impossibility of evaluating subjective judgement in a gen-
eralized way because of its irremediable specificity and context-dependency, the
research on debiasing does not yield any results that can really be generalized and
properly applied to a very wide spectrum of biases, some of its major findings can
be summarized as follows:43
1. Biases mainly arise from the mismatch between the deterministic proceeding of
the human mind and the probabilistic nature of the tasks.44 When the individuals
are asked to assess inferences, they typically rely on subjective procedures and
rules of thumb (the bounded rational heuristics) instead of on the laws of logics
and probability theory. The discrepancy between the probabilistic structure of the
environment and the non-probabilistic solving procedures applied can mainly be
made responsible for the systematic occurrence of mistaken judgements.
2. Under conditions of uncertainty the most delicate phase of decision-making is
represented by the information interpretation which is necessary for structuring
the task, rather than by information collecting. Biases are therefore most likely
41 Cf. Fischoff (1982b, p. 440).
42 Cf. Fischoff (1982b, p. 442).
43 Cf. Fischoff (1982b, pp. 442, 443).
44 Cf. Fischoff (1982b, p. 442).
to be concomitant to the (selective) interpretation of information. On this basis,

the fact that in experimental settings the respondents can typically avoid the col-
lecting phase, as information is typically provided, does not constitute a serious
shortcoming of the experimental method and does not detract from its validity as
a precious instrument of the debiasing research.
3. This can be related to the fact that biases can mainly be ascribed to the individ-
ual approach to the task. They are thus shaped by the interaction between the
individual mental representation, accessibility of the task and its informational
structure. The impact of biases can be reduced by acting on these variables, so
that biases can be said to be non-substantive.45
4. Normative theoretical standards can be inferred as a benchmark for correct in-
ferences. It follows that, when inference assessment is required and given the
structure of the environment, an optimal response to which actual individual re-
sponses can be compared exists.
5. Debiasing only makes sense as long as it exclusively acts on individuals’ intu-
itive judgements. Correcting procedures should therefore strive toward the im-
provement of the quality of subjective judgements by inducing the individuals to
discover them as more favourable patterns of behaviour and to recognise them
as such. Providing responders with external computational aids of various na-
ture (e.g. calculators, computers or similar) does not belong to the instrumental
repertoire for debiasing inference judgements.
6. Biases are meant to have a cognitive and not motivational nature. It means that a
systematic mistake can be defined as a bias only if it can be alleged to shortcom-
ings in the individual judgements rather then to incentives’ or tasks’ triviality.
From these findings it is possible to infer some considerations that can inspire and
inform some of the procedures and methodological fundaments for the analysis of
the self-referentiality of economic theories and in particular for what the testing of
the absorbability of the so-called “mainstream theories”46 may concern.
1. The first finding from debiasing points to one of the difficulties of the analysis
of the absorbability of mainstream economic theories among bounded rational
individuals. This is constituted by the discrepancy between the neoclassical ra-
tionality standard, which underlies such theories and the observable cognitive
limitations which constrain the subjective rationality. As the mismatch between
the individuals’ mental tools and the nature of the task can often be made re-
sponsible for inadequate and biased behaviours, the mismatch concerning the
rational standard respective of the mental tools which actually belong to the sub-
jective cognitive repertoire and of the tools which would be required for properly
conceptualizing and correctly estimating the theoretical support provided by the
theory could interfere with the absorbability of that theory.
2. Neoclassical theories typically privilege the information integration rather than
its discovery. Although neglecting the phase of the search for information
45 Cf. Fischoff (1982b, p. 442).
46 With “mainstream theories” are here simply meant theoretical paradigma inspired by the neo-
classical approach and assuming the full rationality of the economic actors.
5.4 Concluding Remarks on Debiasing and Some Implications for Theory Absorption 117
undoubtedly represents a limit of the neoclassical approach and seriously af-

fects its descriptive power, insisting on the phase of interpreting and integrating
information, which represents one of the utmost needs of the individuals in
their decision-making, could raise the appeal of theories’ prescriptive and strive
toward their absorbability.
3. The third result from debiasing seems to provide empirical support to
Morgenstern’s intuition that the degree to which a theory will be absorbed
decisively depends on its formulation and accessibility.
4. As the absorption of a certain theory points to the evaluation and acceptance both
of its predictive and of its prescriptive content, it results in this way in testing
the survival of a theory to its own acceptance (as can be expressed by the indi-
viduals’ compliance with it). The problem of how bounded rational actors are
able to process the content of theories of full rationality is worth being explicitly
analyzed and considered even when looking for its possible application as an in-
strument for enhancing efficient economic advising. Normative prescriptions for
economic advising could also be taken from theories of bounded rationality, if
it could be proven that such theories enjoy a broader acceptance by the bounded
rational actors and survive to their absorption.
5. While the strive for debiasing rules out the usage of computational aid for sup-
porting the individuals in their decision-making, the investigation of theory ab-
sorption could make use of similar instruments. The automatic implementation
of theoretical content could represent a useful way of testing the individual will-
ingness of absorbing a certain theory because it would answer the question of
whether the individuals would choose to comply with a theory if they only were
able to.
6. The last result sounds quite reassuring because it states the disposability of the
individuals to “work hard” on the task faced. This implies that they will most
likely be willing to have a look at eventual theoretical information provided and
try to assess it as feasible decisional support or not.
Some preliminary considerations on the absorbability of economic theories of full
rationality could be based on the comparison between experts’ and lay people’s
performance. This could enable researchers to specify if and under which conditions
possessing superior knowledge really constitutes an advantage.
One of the advantages of experts and professionals is their experience on a set
of relatively homogeneous tasks, which enables them to reduce the cognitive efforts
needed and to rely on habitual patterns of solution and routinized procedures. Task-
specific reinforcement might add and imply the definition of a manageable range of
decisional clear-cut criteria that can be in many cases made responsible for on-target
judgements. It should however be remembered that experts are in the rule prone to
the same biases as lay people when they predict quite intricate problems on intuitive
basis.47

In conclusion, debiasing and theory absorption are two research programs that
present some similar features because they both represent an attempt to deepen the
understanding of the human cognition and of the subjective rationality.
Effective debiasing requires the development of procedures which act in the
“proper way.” Similar procedures should be accessible to the human mind and affect
judgment in the wished direction by acting on the real determinants of judgement.
Otherwise, they would not permit the systematic blockades of the intuitive judgment
to overcome. In this sense, both debiasing and theory absorption act where cognition
experience fails. They also both insist on the outward boundaries of cognition.
Besides serving the pure speculative intent in exploring the psychology of choice,
these two research programs could yield useful application e.g. in supporting the
development of effective decision support systems, teaching methods, training pro-
cedures, giving policy advice etc. They could also enjoy a broad acceptance among
bounded rational decision-makers.
Both the inquiry of debiasing and that of theory absorption emerge from balanc-
ing theoretical and empirical instances: “Good practice will require better theory
about how the mind works. Good theory will require better practice, clarifying and
grappling with the conditions in which the mind actually works.”48
48 Cf. Fischoff (1982b, p. 444).

Chapter 6
Self-Referentiality of Economic Theories
and Theory Absorption
Bounded rational social actors are not just stimulus-response machines but com-
plex beings, whose actions are led by their own beliefs and mental representations.
Such representations can shape mental models, subjective theoretical frameworks
that predict the course of the social system the actors are involved in and that estab-
lish cause-effect relations that the individual uses in her decision-making. Individ-
uals can modify their mental models when they are not satisfied with the results of
their application. This operation requires that the individuals are able to reflect the
theoretical statements on themselves and on the situation they are confronted with.
According to the result of this reflection process, they will then decide which theory
they want to refer to, or in other words, which theory they want to absorb.
Economic theories aim at the description and prediction of economic behaviour
and interactions, but at the same time interfere with the phenomena they aim to
depict. Revealed theories, if accepted, may influence the behaviour of the agents
they focus on, either in the sense of validation of the theoretical content, or in that
of its rejection.
This analysis tries to discuss the implications of those recursive, or self-reflexive
effects of economic theories on bounded rational economic behaving and interact-
ing. In particular, a distinction will be made between the perception of the self-
referentiality of a theory by bounded rational individuals (i.e. the perception of its
applicability to a concrete setting) and its absorption (i.e. the compliance of the
decision makers with the prescriptions of the theory).
The discrepancy between the neoclassical rationality standard and the observable
cognitive limitations that constraint the subjective rationality further complicates the
evaluation of the role of mainstream theories in influencing economic interactions
and behaviour. The problem of how bounded rational actors process the content of
theories of full rationality undoubtedly is worth being explicitly analyzed because it
could particularly yield interesting results e.g. for enhancing efficient economic ad-
vising. However, normative prescriptions for economic advising could also be taken
from theories of bounded rationality if it could be proven that such theories enjoy a
broader acceptance by the bounded rational actors and survive to their absorption.

120 6 Self-Referentiality of Economic Theories and Theory Absorption
The different approaches to the rationality of the individuals facing economic

problems develop two almost opposite views of economics: while, on the one hand,
the neoclassical teleological orientation codifies economics as a sophisticated the-
ory of reasoning about action; on the other hand, the second view which insists on
evolutionary (in the sense of adaptive) elements as the key for interpreting the hu-
man approach to economic problems, permits theorists to characterise economics as
an empirical behavioural science.
Merging teleological with evolutionary aspects of analysis is a task which deeply
challenges economics. Each of these set of aspects implies a profoundly differ-
ent approach to the economic issues and therefore, yields substantial differences
both in the formulation of economic theories and models and in the assumptions
in regard to their absorbability. While according to the neoclassical approach, a
strong form of theory absorption should be unquestionable; focussing on the bound-
aries of individual rationality allows interpreting theory absorption in a more adap-
tive way.
In the discussion of the absorbability of economic theories, the assumptions on
which such theories rely have to be considered. Therefore, this chapter will be
introduced by a short digression on established economic methodology. The con-
cept of “theory absorption” will then be defined and discussed both in the spirit of
the neoclassical theory and of the bounded rational analysis. The contributions of
Morgenstern (1972), Morgenstern and Schwödiauer (1976) and Dacey (1976, 1981)
will be considered in the case of neoclassical theory while the pioneering work of
Güth and Kliemt (2004a) will illustrate the bounded rational analysis. After some
tentative conclusions, economic policy advising will be discussed as one of the pos-
sible applications of the research on theory absorption.
6.1 Economic Methodology
This short digression discusses some fundamentals of economic methodology and

presents (without claiming to be exhaustive) some of the most established conven-
tions and methodological principles that are widely used and commonly accepted
in economics. For a systematic critical overview on economic methodology, Hands
(2001) can be suggested.
For simplification, the scientific method can be characterized either as based on
an empiricist or rationalist conception and philosophical approach to the human
knowledge.
While empiricism contemplates knowledge as something insolubly depen-
dent on empirical evidence, rationalism interprets reason to be the main source
ofknowledge.
In an empiricist view, any scientific conceptualization must be derived from indi-
vidual direct experience and observation. On this basis scientific hypothesis can
be developed by means of induction, which means that general conclusions are
6.1 Economic Methodology 121
inferred from particular concrete premises. Because of their higher generality, such
conclusions are however not necessarily in possession of the same degree of cer-
tainty as the premises from which they have been generated and therefore, have to
underlie empirical testing.
On the contrary, rationalism appells reason to be the main source of knowledge or
justification.1 Therefore, in a rationalist perspective reason represents the scientific
criterion of truth while deduction is the key mechanism for deriving knowledge
and making science. Deduction involves a sort of concatenated reasoning, i.e. a
reasoning whose conclusions necessarily follow the stated premises which consists
of previously known facts and issues.
These two different approaches to the philosophy of science are not mutually
exclusive. The scientific praxis actually integrates them and has made use of both
inductive and of deductive reasoning.
The epistemologic conception which underlies the two approaches is that of
naı̈ve realism, which does not question the existence of an ontological truth and
believes in the possibility of asymptotic approach toward it but not of its complete
capture.2
Other epistemologic conceptions question this stance and correspondingly posit
different degrees at which science can penetrate reality, providing therefore alter-
native conceptions both of science and of reality. Realism, instrumentalism and
constructivism are examples of the most influential traditions. From the relativist
position of constructivim, which has been sketched in Sect. 1 of Chap. 3, realism
opposes the view that scientific statements actually resemble objective reality, while
instrumentalism interprets scientific statements as useful instruments for expressing
the human experience of reality but does not necessarily reflect the world as it is. As
pointed out by Dacey, “A (methodological) instrumentalist views a theory T merely
as a device of convience in organizing data, i.e., a theory is a mere instrument.
A (scientific) realist takes a theory literally, adhering to the view that the theory
(more or less accurately) describes reality.”3
Economics investigates what on the basis of human behaviour in economic situ-
ations. For doing that it adheres to the epistemological fundaments of the contem-
porary scientific method. Economic situations are represented by all those problems
and circumstances involving scarcity of resources. In other words, economic prob-
lems are faced whenever decisions have to be taken because of and under given
constraints.
Neoclassical economics typically relies on the fundament of methodological in-
dividualism which explains social phenomena as the result of the aggregation of in-
dividual decisions. From this stance neoclassical economics develops the axiomatic
costruction of the “homo oeconomicus,” interpreting the individual as a rational and
self-interested utility maximizer.
1 Cf. Lacey (1996).

2 For an interesting discussion on the relation between theories and reality see Albert (1972).
3 Cf. Dacey (1976, p. 254).
Following the homo oeconomicus approach, neoclassical economics conceptu-

alizes individual decision-making and strategic interaction in the perpsective of
rational choice. Therefore, it assumes an economy to be populated by rational actors
who are in command of “unlimited ressources to analyze models of their interaction
situation and to draw inferences from those models.”4
Modeling choice as rational implies that: (1) rational actors choose how to act in
a concrete situation on the basis of a model of their own action; (2) rational actors
are perfectly able to ascribe causality of events either as due to their own action or to
factors not belonging to the realm of the own activity; (3) they are capable of inten-
tional and purposeful planning which relies on the anticipation of the consequences
of the planned actions; (4) choices are made and evaluated according to individual
maps of preferences which are represented by complete, continuous, convex, tran-
sitive and independent functions of the subjective utility; (5) the individuals assume
the others to be endowed with the same cognitive tools and rationality standards
they are.5
The model of rational choice is thus based on the counterfactual assumptions of
common knowledge of rationality and rational expectations which solve the ques-
tion of the external validity of the rational choice theory on their own axiomatic
basis. Once full rationality of the individuals, common knowledge of that, and ex-
pectations of rationality conform behaviour are assumed, rational choice theory will
be perfectly absorbable among rational individuals.
Rational theory assumes itself to be known (or better inferable on an educative
basis) by all the individuals, who are all aware of that. None of them will therefore
behave differently than postulated by the rational choice theory. This yields the pre-
dictive accuracy of the rational choice paradigm and accounts for the strong notion
of absorption that can be assumed for it. In other words, rational choice theory justi-
fies itself because of its own axiomatic fundaments and in this sense, resembles the
mechanism of metalogic reflexivity.6 As any deductively correct model can explain
what is stated in its own premises and assumptions, observing the absorbability of
the rational choice theory in real settings is a device for testing the descriptive va-
lidity of its assumptions and its predictive sharpness in relation to the individual
(bounded rational) foresight.
In principle, i.e. without the empirical test of its absorption, an explicatory va-
lidity, rather than an explanative one, can be ascribed to the rational choice theory.
Fully rational individuals are able to infer each of its precepts in an educated manner,
which “thereby ‘explicates’ what ‘(ideal) rationality’ means but it does not ‘explain’
choice-making in any nomological sense.”7
In this sense, discussing theory absorption among fully rational individuals is
equitable to the proof of the theory’s internal validity, i.e. of its logical consistency,
4 Cf. Berninghaus, Güth, and Kliemt (2003a, p. 486).

6 Metalogic reflexivity has been defined in Sect. 4 of Chap. 4.
6.2 Self-Referentiality of Economic Theory and Theory Absorption 123
while relaxing any of the assumptions underlying full rationality (as for example
that of common knowledge of rationality) yields interesting implications for the
theory’s descriptive validity.
6.2 Self-Referentiality of Economic Theory and Theory

Absorption
To depict the fact that a social theory which is known to the actors who interact in a
social system may affect the course of the social system itself, Morgenstern (1972)
introduced the concept of “theory absorption.” Although any economic theory can
potentially be absorbed for the resolution of a concrete problem,8 its absorption
differs from case to case, depending on its formulation, its understanding, and its
acceptance by the members of the economic system, as well as on its accessibil-
ity.9 Furthermore, past experiences and learning may also matter when it comes to
evaluating the absorption of a certain theory.
Among the elements that determine or influence theory absorption, self-reference
plays a preliminary role in the sense that it can be characterized as a prerequisite for
the absorption of a theory. Individuals self-refer a theoretical framework to support
their decision-making and then choose whether to rely on it or not, in other words,
whether to absorb that theory or not.
Thus, theory absorption is a consequence of the self-referentiality of the social
theory. An individual self-refers a theoretical statement and, according to the results
of such reflection, she will then choose either to rely on that theoretical framework,
or on a different one.
In an ideal setting (i.e. populated by unbounded rational social actors) a theory
of rational choice will be absorbed universally, such theory being at the same time
descriptive and prescriptive of the full rational behaviour.
In a real setting, thus populated by bounded rational social actors, things are
slightly different and more complicated. A requisite for a theory to be absorbed is
that it can be understood by the individuals (i.e. it does not overstretch their bounded
cognitive and computational capabilities) and that it can be integrated with subjec-
tive beliefs and mental representations. In particular, a theory can be absorbed if its
content does not violate the normative components of beliefs as well as the beliefs
about the others.
Because the recursivity and self-referentiality of a certain theoretical framework
constitutes the logical premise of its absorbability, the notion of “self-referential
theory” will be first of all specified. Then, the original formulation of the concept
of “theory absorption” will be presented and discussed in its implications for a pop-
ulation of fully rational individuals. After relaxing the assumptions posited to the
8 Cf. Dacey (1976, p. 248).

individual rationality by the neoclassical approach, some issues concerning theory

absorption among bounded rational individuals will be finally considered.
6.2.1 Self-Referential Theories
The prerequisite for the absorbability of a social theory has to be sought in the po-
tentially inevitable self-referring nature of any conceptualization about social phe-
nomena. The notion of “self-referential theory,” which therefore assumes a central
meaning in discussing theory absorption, is however feasible for different interpre-
tations depending on the degree to which a theory’s reflexive implications can be
appreciated.
Some of the considerations that follow can be associated with what has been
previously discussed about the reflexivity of human theorizing and knowledge (cf.
Sect. 8 of Chap. 1). These considerations are further informed by some results and
conclusions from the “sociology of scientific knowledge”10 approach.
Science can be characterized as a social activity which is done in communities.
Therefore, it cannot avoid reflecting some of the paradigms that are collectively held
among a certain community,11 which will at least inform the research agenda. In this
sense, scientific theories do not differ from any other intellectual artefact in that they
constitute of socially held beliefs, which are thus coined by the society and at the
same time coin the society to which they refer and by which they have been created.
Relying on that, the approach of the sociology of scientific knowledge postulates
that there is an “explanatory symmetry between scientific and other social beliefs.”12
This symmetry raises the issue of which theory ought to be applied for examining
scientific beliefs because they specify themselves the fundamentals on which their
own analysis has to be based.
This can be referred to theory absorption, as well, since the absorption of a certain
theory sets in motion a recursive mechanism between the content and the object of
that theory. The reflexivity implied by the back-coupling between a social theory and
the social phenomena it aims to describe, therefore, adds to the reflexivity involved
in the analysis of scientific beliefs and statements.
The back-coupling between the theory and its object is influenced by many dif-
ferent factors (e.g. understanding, acceptance etc.) but is essentially determined by
the perception the social actors have of its self-referring character in the concrete
situation they face.
There can be different notions of self-referential theories, at first because there
is much hidden behind the concept of “theory” (as it has also been pointed out by
the introductory considerations on the scientific and economic method) and then
because of the polymorphism recursive relations may assume.
10 For more on the sociology of scientific knowledge, cf. e.g. Davis and Klaes (2003) and Hands
(2001).
11 Cf. Hands (2001, p. 173).
12 Cf. Hands (2001, p. 173) (italics omitted).
“Theory” can be meant as one or several statements that describe a property, a

peculiarity or, more generally, a certain aspect of a certain object. An economic
theory will be conceived here as a conditional generalization, i.e. as a statement of
the sort: “for every x, if P of x, then Q of x.”13
In the perspective of radical constructivism,14 every sort of theorizing would be
self-referential because of the interdependence between observer and observation.
The approach of the sociology of scientific knowledge yields similar conclusions
in that it questions the view of a scientist standing in a disentangled relation to the
world which constitutes, in the end, the object of her analysis.15
However, in a strict sense, a theory should be said to be self-referential if it
refers to itself, i.e. if it contains sentences or theorems related to the theory itself,
as illustrated for example by meta-theoretical considerations. In this acceptation, a
self-referential theory is a theory which applies to itself or which in other words
constitutes one of the objects belonging to its own realm of application.
In a broader sense, a theory that deals with something that can modify or affect
the validity of its content, also implies a recursive relation and thus can be said to be
self-referential. All conceptualizations, where presuppositions involved in knowing
are sought, belong to this category of recursivity. This implicates a meta-logical
referring and occurs whenever observer and observation are part of the same system.
The social sciences are intrinsically exposed to this sort of recursivity since
they are the product of the reflection of individuals upon selected “facets” of the
individual in her social system. Social sciences aim to describe a system made
by an observer, who inevitably is part of the system observed. Considerations on
meta-logical reflexivity undoubtedly suit the social sciences and indicate their self-
referential nature because the kind of social theory which is known to the social ac-
tors who are interacting in a social system can affect the social system itself. “There
is thus a “back-coupling” or “feed-back” between the theory and the object of the
theory”16 depending on which social theorizing can have both a self-supporting or
a self-refusing impact on the social actors and on the social system it aims to ana-
lyze. As social actors can react opportunistically or opposite to the theorizing about
themselves, every theoretical statement can be either invalidated or reinforced by
the actors’ behaviour, as in a feedback loop.
6.2.2 Introducing the Notion of Theory Absorption
The notion of “theory absorption” was first introduced by Morgenstern (1972) and
accompanied by the following considerations:
13 Cf. Dacey (1976, p. 249).

14 Radical constructivism has been discussed in Sect. 1 of Chap. 3. For an introduction see e.g.
Rusch (1999) and Schmidt (1987).
15 Cf. Woolgar (1992, p. 334).
16 Morgenstern (1972, p. 707).
“Nature does not care – so we assume – whether we penetrate her secrets and
establish successful theories about her workings and apply these theories success-
fully in predictions. In the social sciences, the matter is more complicated and in
the following fact lies one of the fundamental differences between these two types of
theories: the kind of economic theory that is known to the participants in the econ-
omy has an effect on the economy itself, provided the participants can observe the
economy, i.e. determine its present state. [. . .]
However, the distribution of the kind of theory available, and the degree of its
acceptance, will differ from one case to the other. This in turn will affect the working
of the economy. There is thus a ‘back-coupling’ or ‘feedback’ between the theory
and the object of the theory, an interrelation which is definitively lacking in the
natural sciences. [. . .]
In this area are great methodological problems worthy of careful analysis. I be-
lieve that the study of the degree of ‘theory absorption’ by the members of the econ-
omy and the study of the above mentioned embedding relationship will make of all us
more modest in judging how far we have penetrated into the economic problems.”17
Its first appearance in Morgenstern’s 1972 article “Descriptive, Predictive and
Normative Theory” does not reflect the real history of Morgenstern’s research pro-
gram on theory absorption. With this concept Morgenstern addressed one of the
topics that had been a central concern all his life. In this regard, “theory absorp-
tion” can be characterized as a subsuming label and as a “refinement of ideas”18
condensed in a single notion, a conspicuous amount of speculations.
Concerns that the feedback which is established between the economic theory
and its argument can result in the unpredictability of the economic system already
find expression in Morgenstern’s 1928 essay “Wirtschaftsprognose.”19 In the core
thesis of this essay is the consideration that in economics “the very fact of forecast
leads to ‘anticipations’ which are bound to make the original forecast false.”20 In
this sense, economic predictions can only be self-falsifying since, as soon as they
are captured by the agents belonging to the economy, they induce them so that they
opportunistically adjust their behaviour to the forecast.
Speculations in regards to the degree at which agents are aware of a theory, play a
central role in Morgenstern’s discussion of the relations between the economic the-
ory and economic policy, as he discussed in his 1934s contribution “Die Grenzen der
Wirtschaftspolitik.” His 1935 contribution further developed this thematic underlin-
ing of the incompatibility between “Perfect Foresight and Economic Equilibrium.”
Morgenstern did not elaborate a formalized definition of theory absorption but
rather focussed on the plurality of elements which might affect the way the eco-
nomic actors interrelate with a certain theory and which should therefore be consid-
ered for the analysis of the recursive effects of a theory and its absorption.
17 Cf. Morgenstern (1972, pp. 706, 707).

18 Cf. Dacey (1981, p. 111).
19 Some further aspects of Morgenstern’s position on economic forecasting have been shortly ad-
dressed in Sect. 2.2 of Chap. 3.

20 Cf. Margert (1929, pp. 313, 314).
6.2.3 Determinants of Theory Absorption
Discussing the absorption of a theory among the members of an economic system

involves the consideration first of the methodological fundaments of economics,
second of the standards met by the human rationality, and finally of the individuals’
in- and foresight capabilities.
As theory absorption can be essentially characterized as a meta-theoretical oper-
ation, its inquiry should include considerations on the methodology on whose basis
theories are formulated, i.e. on the assumptions and simplifications they rely on.
A problem in this insight is constituted by the embedment of the economic real-
ity in the social reality. While any statement on a certain economic phenomenon
inevitably reflects its abstraction from other aspects of reality,21 “the economic
‘reality’ is never a whole, as in the physical universe: Economic phenomena are
embedded in a political, legal, moral and ideological world from which influences
come, also determining events.”22
Evaluating to which extend a theoretical substitute applies to its pre-theoretical
concept23 is a further difficulty worth being considered in a methodological point of
view. It expresses the degree to which an explication stands for the explicated phe-
nomenon and informs the resemblance of the theory with the object it describes,
as well as its internal coherency. It is hereby sought to appreciate the extent to
which a theoretic model can be truly descriptive of the phenomenon it models. It
can be pragmatically conceived that, in principle, theories may be descriptive of
reality and that from case to case a certain theory “may get separated from reality
either because its assumptions are at fault, have become obsolete, e.g. for reasons
of evolution, or because the concepts used in the description can be replaced by
better ones.”24
Theory absorption depends on the real, actual degree to which a certain theory is
reflexive and thus depends on the individual hindsight, foresight and mental capa-
bilities, so far as those elements may affect the “acceptance” of a certain theory.
In this sense, a theory is feasible to be reflexive for certain individuals if it handles
a setting which can be related to the real situation the individuals face. Such a theory
can then be said to be reflexive for those individuals if they actually perceive it as
being related and applicable to the circumstances they are involved in.
Further, the acceptance of a theory by certain individuals is influenced by their
degree of confidence with it. In order to absorb a certain theory, the individuals have
to appreciate it as a normative benchmark on which to inform their decision-making
and have to consider it convincing.
21 This is in Morgenstern (1972, p. 707), explicitly related to predictive theory and expressed as
follows: “Thus the prediction of economic phenomena implies statements about their separability
(or lack of it) from the other social world of which they are only a part.”
23 For a definition of “explication” see Carnap (1956).
The way in which any descriptive theory can assume a normative character is
ruled by its acceptance, as the “normative property is based on the acceptance of
the theory.”25 According to Morgenstern, there are essentially two points which
determine the acceptance of the normative order proposed by a theory: “The first
point relates to the use of theory: if the individual, for example expresses the desire
to behave ‘optimally’ in a specific situation or environment, the theory will tell him
how he ought to behave. The individual will then follow this advice if the theory
is ‘absolutely convincing’. This ‘convincing’ means the intellectual (and practical)
acceptance of the theory for its predictive worth. The individual or agent [. . .] has
to understand the theory in order to develop this degree of confidence. Only then
will he allow his behaviour to be guided, and possibly be changed by the theory.”26
The second point refers to the fact that the acceptance of the normative of a theory
“involves the acceptance of institutions.”27
Foresight capabilities influence the acceptance of a theory and therefore, rule
its absorption, since purposeful action depends on the subjective predicting capa-
bilities. As Morgestern (1935) discusses, the assumption of perfect foresight might
invalidate theoretical equilibrium predictions. This especially applies to those cases
where the equilibrium predictions do not resist their absorption, i.e. when they are
based on a self-destroying dynamic. In similar cases, only admitting individuals
who are not perfectly forward-looking (i.e. not fully rational) could account for the
equilibrium stability.
Finally, the acceptance of a theory can also be affected by individual hindsight,
that is, by the shadow of the past. Morgenstern considers that the formulation of a
theory, its understanding and accessibility by the members of an economy, “[. . .]
will be variously distributed as far as particular theorems are concerned as well as
regards concrete circumstances of the economy in question.”28
6.2.4 Some Theoretical Implications of Theory Absorption
Neoclassical economics assumes theory absorption in its strongest form, i.e. as

full theory absorption. This is allowed because of the neoclassical interpretation
of choices as emergent from the individual rational pursuit of ends with common
knowledge of that.29
From the perfect teleology of action, which is implied by the neoclassical para-
digm, follows that the absorption of a theory can be conceived as a process which is

28 Cf. Morgenstern (1972, p. 707). The text continues as follows: “For example, the population of
a government of a country having just gone through a period of inflation – and therefore having
accumulated experience – will react differently to a new inflation than a country having had no
such recent experience. [. . .].”
29 Cf. Güth, Berninghaus, & Kliemt (2003b, p. 2).
steered by induction.30 Rational actors are defined axiomatically and can therefore
only act rationally, so that the “teleological theory explaining the actions of rational
actors is commonly known and is in fact applied by them in choosing their own ac-
tions.”31 Therefore, rational choice theories consider themselves to be inferable on
an educative basis by every (rational) actor with common knowledge of that. Thus,
in a neoclassical perspective, theory absorption can be characterised as an ultimate
test of the validity of economic theories.
Since a theory of rational choice should constitute a rational standard of be-
haviour, it should survive its own acceptance and not reveal itself as being self-
defeating. As Morgenstern and Schwödiauer (1976) point out: “One of the main
requirements of a satisfactory, or at least acceptable, concept of solution (i.e. ratio-
nal behaviour) is that it ought not to be invalidated by the knowledge of the theory
on part of the participants in the game – it should be immune against ‘theory ab-
sorption’.”32
On this basis they criticize the interpretation of the core as an equilibrium solu-
tion for a market game, arguing its non-absorbability. The absorption of the core as
the set of competitive equilibria would imply collusive re-contracting against which
the core would not resist. Therefore, it would loose its equilibrium predicting va-
lidity because the “knowledge of the core on part of the traders may result in a
collusive stabilization of some dominated imputations.”33
The notion of the core of a market game is the set of non-dominated price-profiles
and its introduction can be traced back to Hedgeworth’s oligopoly solution of the
“contract curve.”34 Competitive contracting should come to an end in the core be-
cause the profit maximizing traders won’t find any trading arrangement which yields
higher profits than those belonging to the core. They will thus agree on one of the
core’s price profiles.
As is the usual case of behavioural theories, the core does not survive its knowl-
edge and acceptance as equilibrium prediction. It is, in other words, not immune to
its absorption, in the sense “the traders are rational – in the sense of always striv-
ing for higher profits – and if they knew that the core were the only stable outcome
of the bargaining process, then the core would not be stable.”35 Knowing that the
core represents the equilibrium prediction would in all likelihood induce rational
profit-seeking traders of the same type to collude, that is, to pre-contract and com-
mit to otherwise dominated agreements. This implies, that “competition can only
prevail if the behaviour of the traders is characterized by a peculiar mixture of ra-
tionality, complete information about the opportunities and the market offers, and
short-sightedness.”36 As this clearly violates the rationality assumptions on which
30 Cf. Morgenstern (1972) and Dacey (1976, 1981).

31 Cf. Güth, Berninghaus, & Kliemt (2003b, p. 2).
32 Cf. Morgenstern and Schwödiauer (1976, p. 218).
34 Cf. Hedgeworth (1981).
36 Idem.
the model of the market is based, the prediction of the core as feasible equilibrium
is a self-defeating prophecy. It constitutes therefore “an example of a theory of ra-
tional action the knowledge of which (on part of the actors) destroys its predictive
validity.”37
The argument for the stability of the core as the set containing the equilibrium
brings into question its suitability as rational standard of behaviour. The alternative
solution concept proposed by Morgenstern and Schwödiauer is that of a stable-set
solution, as conceived by von Neumann and Morgenstern (1944).
A stable-set solution can be defined as a closed set of price profiles which do not
dominate each other (“internal stability” condition) but which can be dominated by
some profiles not belonging to it (“external stability” condition).38
While the property of internal stability applies to both the stable-set and to the
core solution (the different price vectors belonging to the core indifferent to a ratio-
nal trader), the property of external stability is a prerogative of the stable-set solution
because it is not always granted by the core. The strength of the stable-set solution
and its sustainability as a rational standard of behaviour is that it “provides a con-
sistent and defensible rule of division not only for the case of competitive behav-
iour (the core is always part of the solution) but also for all conceivable collusive
arrangements.”39 The stable-set is immune to its absorption because its knowledge
and acceptance as a solution concept would not violate its predictive validity.
Therefore theory absorption can be also considered to be a tool for testing theo-
ries of full rationality at their internal consistency and stability of equilibrium pre-
dictions. The full absorbability of a theory represents a necessary condition for a
theory to represent a rational standard of behaviour in a neoclassical perspective.
In this perspective, testing the theory’s absorption is an operation which essentially
involves the logic of a second-order conceptualization, as it questions how rational
individuals would behave once they become aware of a certain theory.
Relaxing the full rationality assumption, i.e. allowing for boundaries in the indi-
vidual rationality, testing the absorbability of a certain theory (both of bounded and
of unbounded rationality) yields different results. It reveals that if bounded rational
decision-makers perceive that certain theory to be reflexive (i.e. to be applicable to
the situation faced), if that theory induces the individuals to modify their behaviour,
and if it is considered to represent a normative standard, the individuals would like
to conform to it.
However, the implications the enquiry of theory absorption may yield also de-
pend on the assumptions that are chosen to model the logic through which individ-
uals make inferences.
Dacey’s contribution centres on this aspect and explores which consequences the
choice of the logic of inference may have for the absorption of a theory from a
philosophical point of view.
37 Idem.
Dacey formalises the absorption of an economic theory as an inductive accep-

tance process resulting from the maximization of epistemic utility functions. In
doing that, he leans on the philosophical approach which explains the process of
scientific inference by means of decision-theoretic concepts40 and interprets “the
acceptance and rejection of scientific hypothesis as a process of maximizing epis-
temic utilities.”41 Epistemic utilities “represent preferences over cognitive objectives
of scientists, for example, truth, information (in the technical sense of ‘amount of in-
formation’), explanatory power, and simplicity.”42 Dacey applies this notion to the
individual choice of absorbing a theory and assumes this choice to be based on in-
dividual cognitivist inductive logic. This is defined by the pair I = P,U , whereas
P is a measure of the inductive probability43 associated with certain sentences of
a language44 (which consists of conditional generalizations and represents in the
end a theory) and U indicates an (expected) epistemic utility function.45 The (ex-
pected) epistemic utility function should provide a measure of the logical content of
a generalization g relative to the evidence e, and can be therefore specified as
U (g |e ) = P (g |e ) − p (g) (6.1)
A theory T is a sentence of the language over which P is defined if it introduces a

new concept to that language.
Dacey’s analysis demonstrates that the test of a theory at its absorption is in
principle never possible. He corroborates his thesis in that he formally discusses
“three variants of absorption: absorption on the basis of factual evidence alone
and both instrumentalist and realist absorption on the basis of factual evidence and
new (theoretical) conceptual evidence.”46 Whereas the first variant seems to point
to theory absorption in its most neoclassical acceptation (rational actors interacting
with other rational actors cannot act but rationally and cannot expect but rational
behaviour of the others), the remaining two variants emerge as a consequence of
assuming that different individuals rely on different logics of cognitive induction.
This difference can be formally stylized by introducing different functional
specifications of the epistemic utilities, each of which expresses a different
Weltanschauung toward the cognitive elaboration and processing of additional
knowledge. Thus, he distinguishes between two forms of cognitivist logics, the
instrumentalist and the realist ones. In particular, if a theory T is conjoined with
the evidence e, an individual may incorporate it into her epistemic utility in two
different ways, which are:
U1 (g |e&T ) = P (g |e&T ) − P (g) and U2 (g |e&T ) = P (g |e&T ) − P (g |T ) (6.2)
40 Dacey mentions e.g. the works of Hempel (1960, 1962), Levi (1967a, 1967b).
41 Cf. Dacey (1976, p. 250).
42 Cf. Dacey (1976, p. 250, 251), in reference to Niiluoto and Tuomela (1973).
43 For more on the properties of inductive logic see e.g. Hintikka, Suppes (1966).
44 Moreover, the language is assumed to be a monadic, first-order language. See for more Dacey
(1976, p. 251; 1981, p. 116).

45 Dacey relies on the notion of epistemic utilities as introduced by Hintikka and Pietarinen (1966).
Cf. Dacey (1976, 1981).

46 Cf. Dacey (1976, p. 254).
The adoption of U1 rather than U2 reflects the individual philosophical stance on the
relation between theory and reality. In particular, “U1 seems fitted to the position
of a (methodological) instrumentalist, whereas U2 seems natural for a (scientific)
realist.”47
Each of the two different cognitivist inductive logics informs an alternative logic
for the absorption of a theory which can either be an instrumentalist or realist logic
of absorption.
The practical relevance of Dacey’s elegant formal construction can be argued to a
great extent. The questioning starts with how far a process of maximization of epis-
temic utilities may contribute to the better understanding of the individual approach
to theories for the solution of practical problems. Dacey’s analysis brings about the
attempt of explicating (what in a neoclassical spirit has possibly been too often in-
terpreted as equivalent to formalizing) Morgenstern’s notion of theory absorption.
By analysing some of the implications theory absorption yields, economics and eco-
nomic theorizing, he discusses the possibility of the testability theory absorption.
Dacey further investigates several of the factors which result, at a purely theo-
retical level, in the uncontrollability of the absorption of a theory, focussing, among
other things, on the disruptive consequences of inexact (ambiguous) information.
Overall, his analysis “suggests that the concept of theory absorption leads to major
problems in the methodological foundations of inference as well as in the founda-
tion of the social sciences”48 and “reinforces Morgenstern’s recommendation for
modesty on the part of the economists (and social scientists generally)”49
6.3 Theory Absorption among Bounded Rational

Decision-Makers
Several attempts have been made to reduce the bounded rationality approach to the
homo oeconomicus approach, mostly relying on the consideration that the social
actors, though not fully rational, act “as if” they were. Among others, the generally
accepted view that “The aim of a good theory is prediction and in prediction lies
the ultimate test of validity”50 has added some plausibility to the “as if” argument.
But, in spite of that, the fact that in simple settings bounded rational best replies
may coincide with optimal responses, is far from endorsing the “as if” approach.
Moreover, there is no evidence that satisficing behaviour can simply be read as a
step towards optimization.51
Furthermore, it can be argued whether the proof of a behavioural theoretical
framework, thus its validation, can solely rely on its accuracy of prediction, instead
47 Cf. Dacey (1976, p. 254), italics omitted.

48 Cf. Dacey (1981, p. 133).
49 Cf. Dacey (1976, p. 248).
50 Cf. Morgenstern (1972).
51 For more cf. Güth and Kliemt (2004a, p. 522, 523).
6.3 Theory Absorption among Bounded Rational Decision-Makers 133
of also trying to describe the way a decision emerged. In other words, the validity
of a behavioural theory should be proved “from inside,” and in this sense the proof
of the absorption of a certain theory could also be interpreted as an ultimate test of
the validity of such a theory in a concrete setting. As stated, the prerequisite of the-
ory absorption is a self-application of a theoretical framework. Its conditions are its
coherency with the subjective mental representations. The fact that individuals ab-
sorb a theory implies that they accept that they will rely on it - i.e. that they are first
able to conceptualize it in order for them to share the rationality standard underlying
such an approach - and then that they adhere to its logic. A theory that passes such
a test can be said to assume a real descriptive validity of the behaviour of “human
beings” dealing with economic decisions and not just of stylized economic subjects.
A theory is said to be absorbed by an individual if that individual internalizes
it in her own mental models and chooses to act according to its logical content. In
interactive contexts, theory absorption will also be strongly related to the supposed
mental models of the others. It can be distinguished among unilaterally-, partially-
and fully-absorbable theories, depending on the number of individuals - from one
to all - who follow its prescriptions and are satisfied with the result.52
6.3.1 Individual Absorbability of Theories
A theory can be said to be unilaterally absorbable by a bounded rational decision-

maker only if that individual considers herself to be in possession of the theory
and chooses to comply with it, will be satisfied with the theory’s predictive and
prescriptive.
This means that both the extent to which future developments are contemplated
and predicted by the theory and the prescriptions that are claimed, meet the individ-
ual aspiration level and can therefore inform choice and decision-making. In other
words, a theory is individually absorbed if for that individual who absorbs the theory
there is no need or reason for deviating from the theoretical prescriptions.
Because of the teleological perspective underlying any theoretical frameworks,
predictive and prescriptive components of a theory are tightly related to each others.
Both of these components matter for the absorbability of theories in that a purpose-
oriented and forward-looking individual can be reasonably expected first to prove
the theoretical predictions in principle meet the aspiration level and eventually to
choose to behave according to the theory’s prescriptive.
As the prescriptive content of a theory associates certain individual behaviours
to certain consequences and future developments, the mechanism of individual the-
ory absorption can be decomposed into the following steps: (1) only the individual
who is in possession of the theory considers the problem she faces to belong to the
theory’s domain of application, which means, the individual perceives the theory as
self-referring; (2) the theory’s predictive, in principle, fulfils the individual aspira-
52 Cf. Güth and Kliemt (2004a, p. 523).

tions; (3) the corresponding behavioural prescriptions are manageable to the extents
of the individual’s subjective rationality. Whenever this holds, (4) the individual will
choose to rely on the theoretical framework and will thus comply with the theoret-
ical prescriptions. If the individual is satisfied with the feedback she receives from
the compliance with the theory, her behaviour will definitely stick to the theory’s
prescriptive and won’t deviate from it.
A theory can be individually absorbed when its absorption exclusively depends
on the individual who is in possession of it. In particular, when deciding about ab-
sorbing the theory, the individual does not have to build beliefs about other indi-
viduals who interact. Assuming the individual to be fully rational, not only theories
of individual decision-making can be individually absorbed, but also those theories
prescribing dominant strategies in strategic interaction settings which grant such
possibilities. For example, in an ideal setting where a strict dominant strategy exists
for a certain individual, a theory which advises that the individual choose such a
strategy profile must be absorbed by that individual.
However, as soon as bounded rationality is admitted, things become slightly dif-
ferent, because the “practical value of advice to a world of rational being is quite
precarious in a world of less than fully rational individuals.”53 Hereby, the indi-
vidual absorbability requires in addition that the theory’s prescriptive has to be per-
ceived as “good advice.”54
In trying to make the notion of “good advice” more operative, considerable dif-
ficulties are encountered. Good advice should sound “reasonable” to bounded ratio-
nal individuals, should be workable, processible and understandable. This implies,
in particular, that it does not have to overwhelm their limited cognitive and compu-
tational capabilities.
A theory is absorbed in a real setting if it is apt to guide the individual in her
decision-making and inform her choice, meaning that the theory’s advice can be
considered as a standard of behaviour. This can be brought into action only if it
does not overload the individual computational capabilities, which means it is un-
likely that a too sophisticated theory will be absorbed. For example, simple heuris-
tics such as “Take the Best”55 or one-reason decision-making56 in general could be
favoured as stopping rules for making inferences to the rational choice prescription
of extending the search until the marginal benefits of information acquisition are
perceived to exceed the marginal costs.57
Theory absorption could hereby represent a criterion of theory selection, whereas
if the inquiry encompasses both theories of bounded and of unbounded rationality,
uniqueness of theory selection and absorption might not hold.58
53 Cf. Güth and Kliemt (2001, p. 4).

54 Cf. Güth and Kliemt (2001).
55 Cf. Gigerenzer and Todd (1999).
56 Idem.
57 Cf. e.g. Stigler (1961).
58 This motivates, among else, the choice of focussing on testing the absorbability of rational
choice theory in the experimental part of this study (see Chap. 7).
Following Güth and Kliemt (2004a), some related aspects of individually ab-
sorbable theories can be illustrated considering the absorbability of routine behav-
iour in the secretary problem.
In the secretary problem59 an individual has to select one of different items that
are presented sequentially, whereas rejection is an irrevocable choice.
The standard formulation for this problem of optimal stopping is the choice of
hiring a secretary: interviewing as many candidates as possible is useful for collect-
ing information but is a costly process. Adding to that, waiting too much to decide
whom to employ could imply that the best candidates be hired by someone else.
This illustrates a task, in which in each of the N (≥2) rounds, the uppermost
of a staple of N (≥2) randomly shuffled cards, each of which yielding a different
monetary value, is shown to a subject. In each round, the subject can choose between
accepting the card for gaining the monetary reward specified on it or rejecting that
card, so that a new card will be turned. In the final round, the monetary value of the
last card will be paid to the subject.
According to the theory of rational choice, an individual facing a similar task
should first specify a distribution for the values the cards may assume and then de-
termine an optimal stopping rule. This represents a quite demanding procedure for
lesser mortals who could more likely inform their decision-making to the forma-
tion of aspirations and to the search for a satisficing option. It is further plausible
to assume that aspirations can be adapted, in particular, as more final rounds are
approached. Proving the absorbability of this routine of bounded rational behaviour
would corroborate its coherence with the individual mental models and with the
bounded rational search stopping procedures.
6.3.2 Full Absorbability of Theories
In interactive settings a theory can be said to be fully absorbable if all the actors
choose to follow the theory and are satisfied with the results of its application, so
that anyone of the actors have a reason for modifying her behaviour or the theory to
comply with.60 This, in particular, requires that the fact that all actors comply with
the theory does not per se provide an argument for deviating from its prescriptions.
A fully absorbable theory has therefore to admit an equilibrium which is behav-
iourally sustainable among a population of bounded rational individuals. Thus the
theory has to state the “minimum conditions for reaching an inter-personally sus-
tainable ‘reflective equilibrium’ [. . .].”61
In principle, all theories of rational choice should acknowledge their full absorp-
tion since by assuming rational expectations they grant the self-reinforcement of the
59 Cf. Bolle (1979).

compliance with their prescriptions (“rational actors expect the theory to be true
and by behaving accordingly make it true.”).62
In addition to the requirements necessitated by individual absorbability of the-
ories, full absorbability requires that the expectations concerning the behaviour of
the others conform to theoretical predictions.
However, among a population of bounded rational actors it can be argued, on the
one hand, whether considering a theory which prescribes full rational actions it is
reasonable for an individual to expect all of her counterparts to obey the theory and
behave in a rational way. On the other hand, a theory of bounded rationality can
be fully absorbable only if it is apt “to describe how actors are influenced by the
theory itself and how they expect others to be so influenced etc.”63 In this sense, full
absorbable theories have to be reflexive and able to contemplate their own reflexive
implications.
This is another way of looking at the dilemma challenging economics and its
methodology: even if, on the one hand, the solution proposed by the neoclassical
approach on basis of the assumptions of full rationality and its common knowledge
are not sustained by the evidence, on the other hand, the economic theories, which
assume individual bounded rationality, are obviously not provided on the axiomatic
basis for correcting the anticipated behaviour of the others. A theory of bounded ra-
tionality which relies on “the premise that boundedly rational actors assume other
boundedly rational actors to behave according to the theory and themselves behave
according to that theory”64 could offer a possible solution. It could be experimen-
tally tested, for example, to confirm whether or not this assumption is sustained by
the evidence.
In interactive situations the individuals think strategically. They form beliefs
about the behaviours of the others and then (boundedly) best respond accordingly.
Among fully rational individuals, rational and mutual consistent beliefs lead to
the same rational behavioural patterns which enable an (unique and stable) equi-
librium to be identified.65 The assumption of full rationality can therefore neither
explain the possible heterogeneity of behaviour (as all individuals share the same
rational beliefs), nor its persistency (as sub-optimal behaviours are eliminated by
optimal rational ones).
Beliefs may be non homogeneous because they stem from lots of different, highly
subjective factors.66 They can be mistaken and also modified67 if agents are not
satisfied with the result of their application. The factors that shape the subjective
beliefs cannot be classified exhaustively, essentially because the individual, as a

63 Idem.
64 Idem.
65 The general mechanism through which beliefs are updated is learning, which takes place, when
according to Camerer (2003a) a change in behaviour due to experience can be observed.

66 Among them can be mentioned past experience, knowledge, expertise, social norms and indi-
vidual perception of them, risk propensity and its frame, etc.

67 Cf. Camerer (2003a, p. 265).
social actor, is a non deterministic product of her social history.68 The beliefs about
the others, though also influenced by the individual social history, can be much more
characterised as the product of introspection.
It seems reasonable to assume that the individuals perceive themselves not as
perfect but bounded rational, and that they suppose the others possess the same
cognitive and computational capabilities.69 To form beliefs about the others, the in-
dividuals can use the tool of introspection. This is the only way to try to predict the
others’ behaviour since the bounded rational behavioural pattern cannot be logically
inferred. Obviously, being the expression of bounded rational individuals, introspec-
tion can only be bounded, too. Through it an individual is only able to conceptualize
a finite number of iterations of strategic thinking and will then respond strategically
to what she presumes to be the last iteration of thinking performed by the others. In
this sense, each individual perceives herself as being the most sophisticated.
Anew, a theory is fully absorbable if its predictions, as well as its expected ab-
sorption do not question its predictive content. Full absorbability holds for strategic
equilibria70 by definition, since it is optimal to behave according to the expected
equilibrium. Guessing games exemplify a case in which the correct anticipation of
the others’ behaviour improves a player’s chances of winning more than the knowl-
edge of the equilibrium predictions alone. The absorbability of the guessing game
theory by bounded rational individuals will be discussed in Sect. 5 of Chap. 7 in
more details and related to the main findings of an experimental study.
In some other cases, however, theoretical equilibria predictions could be ques-
tioned on basis of their absorbability. This is, among others, illustrated by theories
which focus on the discrepancies between individual and collective rationality (e.g.
theories on the provision of public-goods and social traps in general),71 as well as
by behavioural theories (e.g. theories of oligopolistic competitive equilibrium).72
For behavioural theories absorbability may not be given: for example, in bilat-
eral negotiations73 a theory, which predicts an agreement at the means of the initial
demands, can be confirmed with some intuition and empirical observations.74 This
would, in all likelihood, inspire outrageous initial demands, to which the theory
would no longer apply.
Further, theories of herding behaviour explain rational basis phenomena accord-
ing to which the rational choice of conforming might lead to collective irrational
outcomes exactly because of the compliance with the theoretical advice, i.e. be-
cause of its absorption. This is, for example, the case of informational cascades,
whose discussion in Sect. 6 of Chap. 7 will be informed by the results of an experi-
mental examination.
68 See e.g. Berg, Dickhaut, and McCabe (1995) and Mistri (1997).
69 There is also evidence showing a clear tendency of overestimating the coincidence between the
own and the others’ motives.
70 Cf. Cournot (1838) and Nash (1951).
71 Cf. e.g. respective Coase (1974) and Rothstein (2005).
72 As discussed in Sect. 2.4 of Chap. 6, according to Morgenstern and Schwödiauer (1976).
73 Cf. e.g. Raiffa (1982).
74 Cf. e.g. Pruitt (1981) and Sebenius (1992).
Overall it emerges that, in bounded rational settings, the common knowledge of

a theory does not eliminate strategic uncertainty. Among bounded rational individ-
uals, the common knowledge of a theory does not imply its absorption, which rep-
resents the necessary condition for strategic uncertainty to be ruled out, providing a
reliable basis for predicting the behaviour of the others.
6.3.3 Partial Absorbability of Theories
A theory is said to be partially absorbable if among the population of n (≥ 2) in-

dividuals, there is a subset m (≥ 2) of subjects whose compliance with the theory
would not provide any reason for modifying the own behaviour or for revising the
theory to rely on.
While manifestly in a full rational setting partial theory absorption cannot hold,
partial theory absorption is conceivable among bounded rational actors, as, for ex-
ample, illustrated by a repeated public-good game with voluntary contribution.75
Assuming full anonymous voluntary contribution and that the subjects are solely
informed of the total number of contributions after each round, the individual dom-
inant strategy is that of non-contribution. It would lead, if adopted by all of the
players, to a Pareto inferior result. A possible stylization for this problem is that of
a repeated n-person prisoner’s dilemma in which for the individual i, contribution
yields a payoff of fi (k), while defection that of gi (k), with k being the number of
contributing individuals in the round (0 ≤ k ≤ n − 1).
If (k − 1) individuals contribute, the payoff of the individual i will be fi (K − 1)
if she cooperates and gi (k − 1) otherwise, whereas fi (k − 1) < gi (k − 1) holds for
each of the n individuals and for every k > 0. From the individual point of view free-
riding is better than contributing for any number of contributors, while the condition
fi (n − 1) > gi (0) states the result of universal non-contribution to be Pareto-inferior
to universal contribution.76
While a fully rational individual would in all likelihood stick to the dominant
strategy of non-contributing because she would expect the others to do the same,
a bounded rational individual could condition her choice to cooperate or not in a
certain round to the number of individuals who cooperated in the previous round.
Such a behaviour could be modelled by a (bounded rational) decision rule ad-
vising the individual to contribute if in the precedent round less than k’ individuals
(1 < k < n) contributed and to defect otherwise. Manifestly, a similar decision rule
would not hold among fully rational actors: full rationality implies symmetry among
individuals and homogeneity in the standards of their rationality.
If a subset of individuals conceptualizes a certain rule and behaves on its basis,
while another subset conditions on an alternative principle, it is conceivable for both
rules to be absorbed, although only partially, i.e. only by a subset of the totality of
75 This example has been taken from Güth and Kliemt (2004a, p. 532 ff).
the individuals. A prescription based on asymmetry among the subjects involved can
only be partially absorbed. Partial absorption implies that for the subset of the ab-
sorbing individuals, there is no reason to change the theory or the behaviour because
the outcomes achieved can be assessed as satisficing.
The decision of contributing instead of free-riding would be conditional on the
predictions of the others’ choice. This could, for example, be inferred assuming that
the number of contributors from one round to the successive remains the same if
there were at least k contributing individuals. A number of contributions equal or
higher than k yields a collective satisficing outcome, so that the theory is able to
resist isolated deviations. In particular, “as long as people do not have a suspicion
that others might bring cooperation levels down they may be willing to cooperate
and may go on to do so voluntarily as long as the results are satisfactory.”77 In this
sense, a solution that could not hold among maximizing individuals could be even
robustly absorbed by satisficing individuals.
However, the stability of theory absorption in its partial form remains a highly
arguable issue. Partial theory absorption can only resist up to a certain threshold
value. In the example, spreading uncertainty about the maintenance of the threshold
can drop the rate of contribution.
Some mechanisms for diminishing strategic uncertainty (e.g. commitments, rep-
utation formation, ideological background etc.) could have a stabilizing effect,
as they could consolidate beliefs on the stability of the asymmetric equilibrium
achieved by partial theory absorption. In other words, a partially absorbable the-
ory “must be absorbed by sufficiently many sufficiently influential individuals who
must remain under its spell and individuals must be assured that sufficiently many
will remain so.”78
As partial theory absorption can only be accounted for by relaxing the conven-
tional common knowledge assumptions. It is definitely ruled out by theories of full
rationality. Ad absurdum, a theory of rational behaviour, which would acknowledge
its partial absorbability but would not fulfil the axioms on which it is based, should
therefore be rejected because of logical inconsistency.
On the contrary, admitting the boundaries of human rationality means to recog-
nize the subjectivity of the aspirations, cognitive abilities and mental models and
therefore, allows the asymmetry some settings may imply to be captured.
Indeed, many real-life situations can more realistically be described allowing
for the asymmetry in the characteristics of the individuals involved. For example, it
frequently occurs that individuals with different skills interact: some of them can be,
for instance, more experienced, smarter or simply have a better theoretic knowledge
of the situation they face.
It could seem obvious and would also adopt the neoclassical paradigm of the
homo oeconomicus that a superior endowment of capabilities (experience, exper-
tise, cleverness. . .) constitutes an advantage. In spite of that, what really matters in
interactive situations is to be able to accurately predict the choices of the others in

78 Idem.
order to be able to best respond to their actual (in the observed case of asymmetry
probably inferior) behaviour. In such situations “superiorly endowed” individuals
could misunderstand what motivates the “less endowed” counterparts and so be-
have in a too clever, but unsuccessful way. Among others, results from the ultima-
tum game in Güth et al. (1982) as well as findings from the “less-is-more” effect79
corroborate this thesis.
For those reasons, individuals who possess superior knowledge, expertise or
capabilities could paradoxically decide not to rely - or at least not exclusively -
on them, but on their common sense, in order to predict what a “representative”
bounded rational individual might do.
6.4 Applications of Theory Absorption to Economic Policy

Advising
Although reflexive phenomena have been widely observed in economics, the analy-
sis of the mechanisms that lead bounded rational individuals to accept and eventu-
ally comply (although in a bounded rational way) with the theoretical prescriptions
is still at its beginning. Bounded rational individuals can only process the content
of a theory in a bounded rational way. A better understanding of the mechanisms
on which theory absorption relies could help specifying bounded rational expecta-
tions. Among other things this could enhance the specification of bounded rational
best replies and economic forecasts and could be used, for example, for training
economic professionals and for economic policy advising80 .
Economic policy advising aims at promoting the settlement for political rational
decision. It does not merely refer to the consultation of politicians by economic
professionals, but can be extended to the activity of advising all actors dealing with
political decisions,81 e.g. citizens (either individually or aggregated), institutions
and organisations.
As discussed in this chapter, the boundaries of the subjective rationality influence
the absorption of a theory at each of the steps involved in the absorption process.
Similar to the economic actors, policy makers are at most bounded rational and
therefore, are only able to process the advice they receive in a bounded rational
way. Absorbable advice should acknowledge both the boundedness of its addressee
and that of the behaviours on which it has to be ruled in order to set effective nor-
mative prescriptions and offer realistic descriptive predictions. When bounded and
full rationality reactions to a certain policy are discrepant, advice that is based on
theoretical models of perfect rationality may prove to be incorrect and may fail the
political purpose it was asked to achieve. For this reason, “Explaining to bounded
rational policy makers on the basis of bounded rational behavioural assumptions
79 Cf. Gigerenzer and Goldstein (2002).

80 For some introductory remarks on economic policy advising see e.g. Franz (2000).
81 Cf. Pies (2000, p. 291).
6.4 Applications of Theory Absorption to Economic Policy Advising 141
why and how certain measures may (or may not) work will render policy advice
more acceptable than conventional advice based on welfare maximization.”82
The traditional economic approach to policy advising blames ideologies without
considering that they are indeed at the centre of policy-making.83 Therefore, the
understanding of the role of ideologies and beliefs should be of crucial concern for
providing truly informative and effective advice.
Approaching policy-making in a cognitive and evolutionary perspective aims at
the analysis of the psychological aspects which rule the perception, interpretation
and cognitive processing of the political actors, as well as the related facets of power,
interests and subjective incentives which often motivate political decisions.84
The resemblance between the cognitive and evolutionary approach to politics
and the bounded rational approach to economics is evident. Making allowance for
the bounded rationality of the political actors means to recognize that “beliefs and
ideologies are at the centre of economic policy-making, and can, therefore, not be
overcome by simple proposing ‘efficient solutions’ to policy problems, nor by fos-
tering ‘solid analytical knowledge’ alone.”85
In particular, there are three main questions which challenge the cognitive-
evolutionary view to economic advising: “(i) What beliefs do citizens and policy-
makers actually hold and how do they contrast with economists’ beliefs? (ii) To
what extent do economic beliefs guide economic behaviour and what happens when
economic and political actors behave according to their ‘deviating’ beliefs? And
perhaps more fundamentally: (iii) How do people acquire beliefs, mental models of
theories, and how do they evolve?”86
Results of theory absorption could enrich such a research agenda, hopefully pro-
viding some evidence of the kind of theorizing which is compatible with the actors’
beliefs and mental models and which, therefore, the actors are more prone to accept
and comply with.
The most widely accepted views on economic policy advising (among them
the philosophic constitutional approach of Buchanan and the economic constitu-
tional approach of Rawl are e.g. worth mentioning)87 typically reveal a decisionistic
stance. In such a “decisionistic approach”88 the role of a professional adviser con-
sists in suggesting efficient means for achieving certain goals and purposes that are
given by the policy-makers, as illustrated by Fig. 6.1.
The decisionistic approach relies on assumptions which can be to a great extent
assimilated to the hypothesis of full rationality of the political actors. Therefore, it
often fails to convey a realistic description of the political process and its propelling
forces.
83 For a survey see Cassel (2000). For a discussion of the effects of economics on policy-making
see e.g. Frey (2000) and Wehner (1995).
84 For more on a cognitive evolutionary approach to policy-making see e.g. Meier and Slembeck
(1994), and Pelikan and Wagner (2003).

85 Cf. Slembeck (2003, p. 128).
86 Cf. Slembeck (2003, p. 137).
87 Cf. e.g. Pies (2000).
88 Cf. Slembeck (2003).
Fig. 6.1 A decisionistic ap-

proach to economic policy POLITICS
advicing (Slembeck, 2003)
Political Political
Goals Decisions
Analysis of Options to
Reach Goals
(choice of policy instruments)
SCIENCE
The decisionistic approach in particular requires: (1) policy goals to be fixed and
specified unambiguously; (2) the existence of a first- or second-best solution and
the possibility of profiling it; (3) knowledge of all pros and cons and their correct
assignment to the alternatives advised; (4) compatibility of the advice with the pre-
vailing ideologies and beliefs,89 allowing for its non-revolutionary essence.
As it is quite evident that reality rarely fulfils such ideal requirements, it can
be argued, whether a similar interpretation of policy advising can lead to the for-
mulation of a really effective and efficient advice. By interpreting the definition of
political goals as an endogenous variable, that is paying more attention to the emer-
gence of the political problems and to the setting of the political agenda, this could
provide the basis for a more realistic view on policy making. Besides, considering
political problems and aims as decisively shaped by individually, collectively held
beliefs could enhance the formulation of more acceptable economic advice.
The cognitive-evolutionary view on economic advice90 draws upon similar con-
siderations, whereas one of its basic tenets is that “problems are not fixed and given,
nor do they exist independent of individual perception. Problems emerge through
the perceptions and interpretations of individuals. With regard to policy-making this
means that problems arise at the individual level in that the individual perception of
problems initiates the political process.”91
This leads to conceive policy advising in a procedural, process-oriented way, as
illustrated by Fig. 6.2.
89 Cf. Slembeck (2003, p. 150).

90 It is here referred to its conceptualization by Slembeck (2003).
91 Cf. Slembeck (2003, p. 142).
6.4 Applications of Theory Absorption to Economic Policy Advising 143
POLITICS
Economic & Political
Regulative Political Decisions
Beliefs Restrictions
Advisory Process
Politicians Political Goals Advisors

and Ressources
Scientific Regulative
Paradigms Beliefs
SCIENCE
Fig. 6.2 A process-oriented approach to economic policy advising (Slembeck, 2003)
In the light of the absorption of certain advice, a procedural approach to policy

advising calls for the need to define an interactive role for the economic advisors.
This is compelled by the acknowledgement of the endogenous nature of goal set-
ting, on the one hand, but also by the admission that not all of the proposed means
have the same probability of being accepted (and thus implemented) by the at most
bounded rational policy-makers on the other hand.
Finally, for an interesting application of how the self-referentiality of economic
theories can be empirically tested Lehmann-Waffenschmidt (2006b) can be re-
ferred to. He infers conclusions for professional advisers on optimal advising in the
case when self-referential effects may occur. Lehmann-Waffenschmidt (2006b) also
presents an application of this idea to the case of underpinned advice which is con-
fronted with delaying reaction behaviour of the addressees. Considering the propen-
sity of the decision-maker to postpone the execution of given advice in dependency
of the perceived urgency of its underpinning argument, Lehmann-Waffenschmidt
(2006b) discusses the objectives of professional advisers as follows: “Naturally, his
first objective is to choose such an advice and such underpinning arguments that
the advice really will be taken by the addressed agents (argument justification ob-
jective). This is closely related to the problem of the predictability of social events
[. . .]. The adviser’s second objective [is that] of being right with his underpinning
arguments and his third objective (is) [. . .] his potential self-interest in the ultimate
outcome.”92
92 Cf. Lehmann-Waffenschmidt (2006b, p. 3).

Chapter 7
On the Absorbability of Economic
Theories – An Experimental Analysis
Although reflexive phenomena have been widely observed in economics, the

analysis of the mechanisms that lead bounded rational individuals to accept and
eventually comply as far they could with theoretical prescriptions is still at its
beginning.
The point is that even if, in principle, the reflexive implications of social theoriz-
ing on the economic actors can never be ruled out, their occurrence actually depends
on the understanding, acceptance and coherence between a certain theory and the
individual mental models. The recursivity of economic theories and their absorption
will therefore differ from case to case, which means that the inquiry of this topic
cannot only rely on theoretical speculations but needs to be supported by empirical
findings. In particular, the analysis has to address the following issues: (1) how real
economic actors perceive the recursive character of economic theorizing; (2) if and
under which conditions economic theories affect the behaviour of the economic ac-
tors in a self-referential way; and (3) how the self-referentiality of economic theories
can be empirically tested.
The experimental part of this study can be conceived as an attempt to approach
the wide and complicated field of the recursivity of economic theories and to explore
the possibility of testing the validity of such theories relying on their absorbability
among bounded rational individuals.
In the first section of this chapter, after some introductory considerations on the
experimental method in economics, a possible approach to the experimental analy-
sis of the self-referentiality and absorbability of economic theories will be proposed.
In this regard, some related research concerning absorbability and task transcending
of satisficing will be discussed. The indicative results of some preparatory attempts
at testing the self-referential implications of theories that have been conduced in
the form of pilot and classroom experiments will be briefly presented. In consid-
eration of their scarce statistical validity and questionable methodological hygiene,
however, extreme caution is advised in interpreting the hints they provide.
The second section will then be dedicated to the analysis of two experimental
studies. The first is focused on the absorbability of equilibrium predictions on guess-
ing games (cf. Sect. 5 of Chap. 7), while the second discusses the absorbability of
146 7 On the Absorbability of Economic Theories – An Experimental Analysis
informational cascades’ theory. Each of the two studies will be discussed by in-
troducing previous related experimental evidence, by presenting the specific design
of the experiment with its main results. Some tentative conclusions which can be
inferred on the basis of the experimental evidence so far conclude the chapter.
7.1 The Experimental Method in Economics
While for the so-called hard sciences (like physics, chemistry and biology) the ex-
perimental methodology constitutes the prevailing technique for inferring and test-
ing theories, the application of the experimental method to the enquiry of economics
represents a clear rupture with the traditional non-experimental approach. With this
approach, economics essentially conceived as a non-experimental subject and posits
accordingly that economic data can only be collected in a passive way, by observing
real developments and phenomena under exogenously given circumstances, i.e. not
under controlled conditions.
Experimental economics, whose first attempts can e.g. be traced back to the
work of Thurstone (1931), Preston and Baratta (1948), Mosteller and Nogee (1951),
Allais (1953), Edwards (1953a, 1953b), May (1954), Davidson, Suppes and Siegel
(1957), Davidson and Marschak (1959),1 shows indeed that economic data can be
generated in the laboratory under controlled conditions. This makes for better qual-
ity data, available at lower costs and in reasonable amounts of time.
The conception underlying experimental economics is “to use the lab as a hot-
house for cultivating theory: Existing lab data guides model construction. Once
constructed, the model is subjected to a new round of tests. Depending on the re-
sults, the model is either refined or abandoned.”2
Experiments can serve different purposes, which have been colourfully labelled
and are well-known as “Speaking to Theorists,” “Searching for Facts,” “Searching
for Meanings,” and “Whispering in the Ears of Princes.”3 They strive respectively
for testing formal theoretical predictions, for studying the effects of variables not
contemplated and/or not fully seized by existent theories as well as for developing
mechanisms of effective policy advising.
The setting of an experiment consists in a controlled economic environment,
whereas in an “economic environment” one or more individual economic agents in-
teract together through an institution.4 Whilst achieving control over the institution
is a straightforward operation in an experimental environment, extending the con-
trol to the agents involved requires the codification of some additional assumptions
and procedures. An essential point in this regard is to strive toward the emergence
of constitutive economic characteristics, such as preferences, technology, resource
1 For a review, see e.g. Kagel and Roth (1995).

2 Cf. Bolton (1998, p. 258).
3 Cf. Roth (1995, pp. 22, 23).
4 Cf. Friedman and Sunder (1994, p. 13).
7.2 Testing the Self-Referentiality and Absorbability of Economic Theories 147
environment and information. For achieving that, experimental economics steers on

the reward medium which relies on the so-called “induced-value theory.” 5
Induced-value theory posits, as sufficient conditions for controlling agents’ char-
acteristics, the three requirements of monotonicity, salience and dominance of re-
wards. They state respectively: (1) that subjects’ preferences should be strictly
monotone which increases the reward medium (which is normally achieved by using
domestic currency as the reward medium); (2) the reward received should be related
to the action taken in a way that is intelligible to the subjects; (3) that, in the experi-
ment, changes in subjects’ utility are predominantly due to the reward medium while
other influences are negligible. These conditions are, as a rule, made operative by
choosing a performance-dependent reward scheme whose average rewards can be
estimated to exceed the subjects’ average opportunity costs. In addition, the subjects
have to be provided with simple, clear instructions, drawn in neutral terms, whose
understanding can be e.g. checked in preliminary dry runs and/or control questions.
Privacy concerning the individual performance and payoffs must be granted, and
the experimental aim must be kept secret. Experimenters cannot deceive or lie to
the participants in an experiment regarding none of the experimental features.
The evidence which is gained from economic experiments can then be trans-
ferred into theory according to the so-called “parallelism precept,” which states that
“[p]ropositions about the behaviour of individuals and the performance of institu-
tions that have been tested in laboratory microeconomies apply also to nonlabora-
tory microeconomies where similar ceteris paribus conditions hold.”6
Experiments have been revealed to be an essential support for economic and, in
particular, microeconomic theorizing. So far, a myriad of experiments have been
conduced, deepening the economic behaviour in almost all microeconomic situa-
tions contemplated by the economic theory. For a survey on milestone-contributions
in experimental economics and a systematic discussion of their main findings and
achievements see, for example, Kagel and Roth (1995) and Hey (1994).
7.2 A Possible Experimental Approach for Testing

the Self-Referentiality and Absorbability
of Economic Theories
The experimental investigation of the self-referentiality of economic theories and

their absorption should include a large spectrum of experiments. As said, this is
because even if absorbability of economic theories is, in principle, always possible,
its concrete modality can only be highly context specific and follows the complex
mechanisms ruling bounded rational cognition and decision-making. Therefore, the
experimental testing of theory absorption should, in an ideal case, be articulated over
a broad range of experimental situations so as to observe how the specific faculties
5 Cf. Smith (1976).

6 Cf. Smith (1982, p. 936).
involved in different scenarios of economic problem-solving and decision-making

affect the perception of a certain theory and eventually create the conditions for its
absorbability.
As previously discussed in Chap. 6, assuming different rationality standards for
the economic actors yields different implications for the recursivity and absorbabil-
ity of the economic theory. Correspondingly, this theme can be approached either
focussing on the reflexive effects of theories, which rely on the assumption of fully
rational economic actors or on those of theories encompassing individuals’ bounded
rationality.
This study conceives the inquiry of the absorbability of theories of full rationality
as a natural first step and is preferred over starting with the observation of the absorp-
tion of theories of bounded rationality. This is inspired by different considerations.
First, as their denomination suggests, mainstream theories enjoy a broad acceptance,
which (with their common axiomatic basis and underlying assumptions) leads to
their almost univocal codification when applied to a concrete setting. Adding to
that, theories of full rationality interpret the individuals as straightforward utility
optimizers. They greatly simplify aims and purposes which steer human decision-
making, which is manifestly reductive but still appealing whenever it comes to eco-
nomic decisions. In principle, mainstream theories can be assessed to be, to some
extents, familiar to the people, not for what concerns their content, but for the logic
inspiring them.
A further aspect, which sides with first focussing on the absorption of full ra-
tional theories, is represented by the ideal normative standards they posit. Although
their descriptive validity can be seriously questioned, their usefulness as a normative
benchmark for assessing decisional and judgemental quality holds in many cases.
On the contrary, theories of bounded rational behaviour do not enjoy such an
ideal normative standard. Despite that, they contemplate a more complex range of
aims and purposes which may motivate the acting individual. This adds to their
plausibility and their descriptive power, but at the same time increases the variables
that should be accounted for.
Moreover, theories of bounded rationality are articulated over a particularly
vast panorama of different approaches that frequently ascribe the responsibility for
problem-solving and decision-making to very different faculties of the human cog-
nition. Because of their pronounced empirical anchoring, bounded rational theories
can be thought of as more realistic in their description of human cognition than
neoclassical theories.
Plenty of experimental evidence reveals how theories of unbounded rationality
are not always descriptive of the observed economic behaviour. Now, the proof of
the absorption of such theories could work as a test of their “viable” normative
validity. On the other hand, as theories of bounded rationality have tried to interpret
and stylize the systematic violations of theories of full rationality, theory absorption
could be seen as an ultimate test of validity of their real descriptive power, as well
as of the degree of their acceptance.
The effects of the self-referentiality of economic theories and their absorption
can be experimentally analyzed through the observation of how individuals deal
with meta-theoretical information in different experimental contexts.
7.2 Testing the Self-Referentiality and Absorbability of Economic Theories 149
Meta(-theoretical) information (i.e. theoretic information about the experimen-

tal situation the experimental subjects face, which aims at adding to the theoretical
knowledge of the subjects and not simply to their information)7 could be commu-
nicated to the experimental subjects in the form of “meta-instructions.” By “meta-
instructions” instructions it is meant, instructions that reveal the theory underlying
the experimental situation and/or previous experimental findings. The information
contained in the meta-instructions should not overwhelm the cognitive bounded ca-
pabilities of the individuals and should be secured by some control questions on the
application of the meta-instructions or supplemented by a questionnaire at the end
of the experimental sessions.
The experiments have been run over two treatments, with and without meta-
instructions. The comparison between the two treatments intends to enhance some
conclusions both on the perception of the self-referentiality of economic theories
and on their absorption. While a significant difference between the behaviour of the
test and of the control groups (i.e. respectively with or without meta-instructions)
can reveal the perception of the self-referentiality of the theory communicated by
the meta-instructions, the compliance of the experimental subjects with the theory
conveys a proof of its absorption.
The experimental hypothesis underlying this study is that meta-theoretical in-
formation can support the bounded rational decision process and improve the out-
come’s efficiency degree in various experimental settings. In the ideal case, the ex-
perimental research should cover both individual and interactive decision-making.8
Meta-instructions can be also interpreted as an attempt to support the subjec-
tive rationality, not in the sense of considering theory as “congealed experience,”9
according to which, experience could be substituted by theoretic knowledge, but
in the sense that if non-optimizing behaviour would simply be a step in a discrete
optimization process (as sustained by the “as if” approach), individuals would
choose to optimize, if they could; i.e. they would comply with the theory presented
in the meta-instructions. It can be argued whether that really happens. In many
situations, as for instance in beauty contests, coordination or public good games,
common knowledge of the equilibrium does not eliminate strategic uncertainty.
In similar settings the outcome of the game cannot be foreseen and the players
are mainly concerned with predicting the others’ behaviour. In economic inter-
actions where bounded rationality of the others matters, meta-information could
be expected to promote the emerging of a sort of meta-rationality, a behavioural-
rationality, which transcends full rationality and is superior to it in terms of success
in concrete settings.
However, the absorption of meta-instructions is not a trivial question even for
unilateral decision making. Supporting the subjective bounded rationality is far from
leading to perfect rationality, as corroborated by the resistance of several cognitive
biases to debiasing attempts as well as by the evidence on “less is more” heuristics.
7 The difference between “knowledge” and “information” is here meant as in Sect. 1.2 of Chap. 3.
8 Except for a trial on the debiasing of the conjunction fallacy through meta-instructions, the evi-
dence discussed in the present study restricts to interactive decision-making.
9 The expression stems from Morgenstern (1972, p. 707).
The absorption of meta-instructions requires their coherency with the subjective

beliefs in order to be trusted as a valid support to the decision. So, even in situa-
tions where it has been experimentally shown that actual behaviour differs from the
predictions of the rational choice theory, theory absorption could test its normative
acceptance and validity.
The compliance with meta-instructions based on theories of bounded rational-
ity could help in testing the real coherency between their underlying assumptions
and the mental models of the individuals. The proof of the survival of a theory
of bounded rationality to its own acceptance could say more about its cognitive
reliability.
In Chap. 6 the self-referentiality of a certain theory has been interpreted as a
logical premise for its absorbability, and its perception as one of its necessary con-
dition. In this sense, the perception of the self-referentiality of a theory could be
disentangled from its absorbability whenever the experimental evidence accounted
for significant differences between the behaviour of the individuals in the absorption
and in the control treatment, on the one hand, and between behaviour and theoretical
predictions on the other hand.
7.3 Related Experimental Studies
The experimental testing of economic theories at their absorbability level is still at

its earliest stages. A research project on the absorbability of satisficing has recently
started by the Research Group on Strategic Interaction at the Max Planck Institute
for Economics in Jena. The experiments that, to the best of our knowledge, had
been conducted in this field, the experimental procedures they adopted and their
main findings will be briefly sketched in the following pages. They aim at testing
the absorbability and task transferability of the satisficing approach in a portfolio
selection task10 and on exploring whether individuals prefer to conform to principles
of bounded rationality rather than to rational choice.11
The first study12 observes the behaviour of individuals in different financial
investment tasks with state-specific return rates and with a finite number of states.
Such states can be ordinally assessed, associated with different degrees of risk and
correspond to different degrees of risk. Different states of nature are equally probable.
Individuals faced with a similar task can be said to behave satisficing if they
reveal consistency of return aspirations (i.e. their aspirations refer to non-empty
portfolio sets) as well as consistency between their investment decisions and the
aspirations they state.
The experiment was articulated over two phases. During the first phase, the satis-
ficing principle of aspirations consistency was enforced by means of a routine. The
10 These aims are respectively tackled in Güth, Levati, and Ploner (2006) and Fellner, Güth, and
Martin (2006a).
11 Cf. Fellner, Güth, and Martin (2006b).
12 Cf. Güth, Levati, and Ploner (2006).
7.3 Related Experimental Studies 151
respondents were asked to form their subjective return-aspirations as long as they

achieved consistency. After that, participants were left free to choose a portfolio and
were provided with feedback regarding whether their choice satisfied their return as-
pirations or not. During the second phase, the respondents could decide whether to
further rely on the satisficing routine or to make their portfolio choice freely, while
having relaxed, in this case, the requirement of aspiration consistency.
In this setting, evidence for the absorption of the satisficing approach would have
been inferable on the basis of the compliance with it. The key chosen to interpret
the results was to assume theory absorption to be confirmed when the individuals
acted satisficing, i.e. in respect of the two principles stated above. Accordingly, a
proof for the absorbability of satisficing would have been during the first phase
when individuals opt for a portfolio which meets their aspirations and therefore,
revels adherence to the logic underlying the satisficing approach. The respondents
who in the first phase absorbed the principles of satisficing were expected in the
second phase either to further rely on the satisficing routine or to refuse it but still
to comply with the principles of satisficing.
The overall findings seem to confirm the individual ability to absorb the satisfic-
ing approach. The data reveal that almost half of the participants who refuse the aid
after the enforcement phase still chose satisficing portfolios and that most routine-
assisted investment choices are consistent with the aspirations stated. Therefore, the
study provides evidence for the possibility of learning the principles of satisficing
and at a degree which is dependent on the tasks complexity.
A seemingly disrupting finding is represented by the occurrence of a suboptimal
routine among a minority of respondents, according to which, participants did not
modify their return-aspirations from period to period.
The second study associates the inquiry of the absorbability of the satisficing ap-
proach with some reflections on the universality of the bounded rational approach.
This is tackled by testing the prescriptive applicability of aspirations formation and
satisficing behaviour with their transferability among similar tasks. For this, the re-
spondents were faced with two investments tasks which differed in their complexity.
The experimental data are analysed by clustering the participants according to
their responses into the three categories of unreasonable, potentially satisficing and
actually satisficing. These clusters indicate respectively if the stated aspirations do
not meet the minimal requirements for consistency, if they define a non-empty set
of investment choices, and if they are fulfilled by the investment choice done.
The distribution of the respondents in the different categories confirms that prin-
ciples of satisficing are not innate in the individuals but need teaching or consulting
and learning. The efficacy of learning could however be interpreted in the sense
of the absorbability of the satisficing approach. There is some evidence hinting at
the high consistency of aspiration formation among tasks, whereas the relative low
number of potential and actual satisficers among the respondents suggests caution
in this insight.
The third study13 finally explores whether bounded rational individuals, involved
in a simple portfolio choice task, prefer to rely on the satisficing or on the optimizing
13 Cf. Fellner, Güth, and Martin (2006b).
approach. In an initial phase the participants were familiarized with two routines,
each of them enforcing one of the two approaches. Routines were automatically
processed on the basis of individual parameters the participants were asked to ex-
press. Such parameters expressed down- and upward aspirations and the subjec-
tive attitude toward risk. In each period and according to the result of the routine
processed, the participants were forced to implement a certain investment. In a sec-
ond phase the participants were left free to decide whether to opt for one of the two
decision modes or to formulate their investment choice freely.
From the analysis of the data it emerges that optimizing was favoured over satis-
ficing by 62% instead of by 38% of choices.
7.4 Some Preparatory Attempts
The indicative results of some trials and pilot experiments, which have been mostly
conduced in the classroom, inspire the tentative considerations that will be discussed
in this paragraph. It should be premised that neither the size of the sample (which
does not offer enough independent observations to corroborate any result), nor the
experimental conditions (in a classroom perfect isolation of the subjects cannot be
achieved) allow one to trust such results as definitive evidence. In addressing what
emerged from these preparatory attempts, it is not intended here to provide facts but
merely tendencies or possible developments that could help in sketching the experi-
mental framework for testing theory absorption among bounded rational individuals.
In particular, pilot classroom experiments will be focussed on that deal respec-
tively with (1) the experimental attempt of debiasing the conjunction-effect bias
through meta-information,14 with (2) the role of the theory of integrative negotiation
in promoting efficiency in multilateral negotiations, and with (3) a guessing-game
with information-feedback and meta-instructions.
As previously put forward, the analysis has been restricted to main stream, hard-
core theories that rely on the assumption of full rational economic actors.
As follows, an overview of the experimental procedures adopted for the different
attempts will be offered, and the main tendencies that can be subsumed for the
observation of their results will be mentioned. The experiments’ instructions are
available upon request.
7.4.1 An Experimental Attempt of Debiasing the

Conjunction-Effect Bias through Meta-Information
To test the absorbability of theoretic principles in non-interactive decision-settings,

a task has been focussed on in which typically the conjunction-effect bias occurs.
14 Referring to the diploma thesis of Pombeni (2005).

7.4 Some Preparatory Attempts 153
This bias consists in the systematic violation of the conjunction rule in subjective
probability judgements and is typically ascribed to the representativeness heuristic.
This fallacy has been the focus of a considerable amount of experimental studies
and has been shown to be quite resistant to debiasing attempts.15
A classroom experiment, which strived for the debiasing of the conjunction rule
through meta-information, was conducted in 2005 on 30 undergraduate students of
the University of Dresden.
The participants had to solve a task involving probability estimation and were
divided into two groups, a control and a debiasing one.
The participants assigned to both groups had to solve the same task, which
consisted in estimating and ordering the probability of some conjuncted and non-
conjuncted events. The only difference between the two treatments was that the par-
ticipants assigned to the debiasing group received theoretical information in form
of meta-instructions containing an explanation of the conjunction rule, while the
respondents belonging to the control group were not provided with such a deci-
sional aid.
The evidence from this experiment was that while in the control group the con-
junction bias occurred by 75%, its frequency in the debiasing group was reduced to
13.3%. This seems to suggest the absorbability of the conjunction rule and hints at
the efficacy of meta-theoretical instructions as a possible debiasing procedure.
7.4.2 A Classroom Experiment on Theory Absorption

in Multilateral Integrative Negotiations
Both distributive and integrative approaches to negotiation are in principle not im-
mune to their absorption; their knowledge would inspire outrageous demands to
which they would, in all likelihood, no longer hold. Their absorbability decisively
depends on the foresight of the contrahents, thus on their rationality and motives, as
well as on the institutions which rule the negotiation.
Even if real multilateral negotiations in many cases offer a potential for inte-
grative agreements, negotiators often fail to exploit the possibilities of maximising
joint benefits and choose sub-optimal bargaining solutions. This trial aimed at ex-
perimentally analysing if meta-information of integrative negotiation theory can be
absorbed from the negotiators and promote the settle for optimal integrative solu-
tions. Specifically, the experiment focused on the integrative negotiation technique
of logrolling which permits the respondents to solve the negotiation on a win-win
basis whenever preferences are asymmetric, by means of the mutual exchange of
concessions.16
In 2003 a classroom experiment was conducted at the University of Dresden.
In the two sessions, 18 undergraduate students were divided into groups of three
15 For more on the conjunction effect see Sect. 3.1.2 of Chap. 5.

16 Cf. e.g. Stratmann (1997).
and had to negotiate some issues. The first session resembled the design used
by Lehmann-Waffenschmidt and Reina (2003), from which it emerged that under
the majority rule the integrative potential of a negotiation does not, as a rule, get
exploited.
In the first session, each group had to negotiate over three issues under the major-
ity rule. Free talk was admitted and individual preferences were asymmetric. In the
second session, groups were rematched and the respondents were confronted with
the more complex task of negotiating six issues. The same rules as in the first session
applied. Besides, in the second session participants were provided with theoretical
information in the form of a decisional aid explaining the principles of integrative
negotiation and logrolling.17
The evidence from this trial hints at the absorbability of the principles underlining
integrative negotiations: whilst in the first session the integrative solution was cho-
sen by 4 of 6 groups, in the second session, despite of the increased complexity of
the negotiation, all groups settle for an integrative solution. However, even after hav-
ing provided the students with principles of integrative negotiation, the integrative
solution that was favoured by all groups was the fair one, yielding the same payoff
for each individual, rather than a very optimal one, corresponding to an asymmetric
payoff distribution, which might have increased the danger of majority cycles.
7.4.3 An Experimental Guessing-Game in the Classroom

with Information Feed-Back and Meta-Instructions
A classroom experiment, which was run in January 2006 among 266 undergraduate
students of the University of Dresden, focused on the role of theoretical information
and previous empirical evidence in a p-guessing game.18
The purpose was to analyse the absorbability of game theoretical principles for
inferring the Nash equilibrium and to compare it with the effects of spreading infor-
mation on empirical evidence. Meta-information, both regarding theoretical princi-
ples and empirical results, was expected to promote the emergence of a sort of game-
rationality, which is among bounded rational actors superior to the game theoretical
prescriptions.
The guessing game represents a setting, in which common knowledge of the
equilibrium does not eliminate strategic uncertainty. The outcome of a guessing-
game cannot be foreseen, as the players are mainly concerned with predicting the
others’ choice. More on the guessing game, its theoretic modelling and experimental
analysis will be, for convenience, treated later in Sect. 5 of Chap. 7 when discussing
one of the core experimental studies of this research.
For this trial the experimental setting of a repeated p-guessing game with p = 2/3
was chosen. Within the close interval [1,100] the participants had to guess a target
number corresponding to the entire number which was closest to the two-third of
17 For more details on logrolling see Mueller (1997).
18 More on the evidence of this experiment is presented in the diplomat thesis of Marx (2006).
7.5 On the Absorbability of Guessing Game Theory 155
the mean of all chosen numbers. The three participants with the closest guesses to
the target yield a reward of 5 e. Three treatments were run: the 89 participants
assigned to the control treatment did not receive any further information except the
instructions, while the 87 participants that took part in the data treatment received
a decisional aid in the form of feedback information about the empirical results
of a previous similar p-guessing game.19 Finally, in the absorption treatment 89
individuals were provided with theoretical information about the Nash equilibrium
and how to derive it through iterated elimination of dominated strategies.
Figure 7.1 summarizes the frequencies of guesses in the different treatments.
The experimental hypotheses were that (1) the target number in the absorption
treatment as well as (2) in the data treatment were significantly lower than that in
the control treatment. While in the absorption treatment, theory absorption could
implement the choice of the Nash equilibrium, in the data treatment, the provided
empirical findings could represent an anchor for the participants to make their guess.
The results of t-test indicate that both first and second hypotheses cannot be re-
jected respective to p = 0.008 and p < 0.001. When comparing the results of absorp-
tion and control treatments, it emerges that the target number in the data treatment
is significantly lower (p < 0.001), i.e. closer to the Nash equilibrium, than in the
absorption treatment. This seems to hint at the incomplete absorption of the prin-
ciples of iterated elimination of dominated strategies20 and can be ascribed to the
strategic uncertainty which is unavoidable in guessing games. In this sense, infor-
mation on empirical evidence might have contributed to reduce strategic uncertainty
by providing an anchor for typing a guess.
7.5 On the Absorbability of Guessing Game Theory21
Besides, evaluating the trial on a p-guessing game confirms the suitability of this
setting for testing theory absorption. The p-guessing game has namely several at-
tractive characteristics for deepening rationality and learning, in that it permits the
effects of rationality to disentangle from social preferences such as inequality aver-
sion, fairness or reciprocity, which are, at the same time, very simple to explain,
and have a very simple economic interpretation. Therefore, it represents a suit-
able framework for experimentally testing theory absorption by bounded rational
decision-makers.
As follows, after a more detailed introduction on the guessing game, an overview
of relevant studies will be offered, including both theoretical and experimental
19 The empirical evidence provided stems from a previously conduced trial, which had been run
under comparable conditions, i.e. as classroom experiment, among undergraduate students of the
University of Dresden, and following the same experimental procedure and modalities. A Welch’s
test proved pooling of the samples to be allowed.
20 This will be discussed in more details in the next paragraph (Sect.5 of Chap. 7).
21 The results of this experimental study and the data analysis are discussed in more details in
Morone, Sandri, and Uske (2008).

(a) control treatment N =89

10
9
8
frequency
7
6
5
4
3
2
1
0
1 11 21 31 41 51 61 71 81 91
guess
(b) data treatment N=87

10
9
8
7
frequency
6
5
4
3
2
1
0
guess
(c) absorption treatment N=89

10
9
8
frequency
7
6
5
4
3
2
1
0
1 11 21 31 41 51 61 71 81 91
guess
Fig. 7.1 Frequency of guesses in (a) the control treatment, (b) the data treatment and (c) the ab-
sorption treatment
contributions. A model for deriving the Nash equilibrium by means of iterated elim-
ination of dominated strategies will be presented and aspects of partial and full the-
ory absorption for guessing games will be related. The experimental design and
hypothesis follow, and the main results and some final remarks on them conclude.
7.5.1 On the Guessing Game
The logic underlying the guessing game was first pointed out by Keynes (1936)
as a simplification for describing the behaviours of investors in financial markets.
As “most of these [professional investors] are, in fact, largely concerned, not with
making superior long-term forecasts [. . .] but with foreseeing changes in the conven-
tional basis of valuation a short time ahead of the general public”22 the logic inspir-
ing the behaviour of professionals on financial markets resembles one of newspaper
beauty contests, which ask respondents to guess which one among several faces will
be appointed as the prettiest by the majority of the newspaper’s readers.
Similarly, in a p-guessing game all n(≥1) players i = 1, . . . , n have to choose
simultaneously a number gi from a closed interval [L,H], with 0 ≤ L < H. With p
belonging to (0,1) and d ≥ 0, the wining player is the one whose guess is closest to
the target number:23
n
gi
g = p ∑ +d
∗
(7.1)
i=1 n
Commonly known, iterated elimination of dominated strategies leads to derive the

unique equilibrium, as the players are able to conceptualize the equilibrium predic-
tion when they all know that all of them know that all are aware of the principles of
elimination of dominated strategies and are capable of deriving its implications.
This illustrates the full absorbability of strategic equilibria in guessing games
when the assumption of common knowledge of rationality holds. In a similar set-
ting, partial theory absorption can be illustrated assuming that just a subset of the
n ≥ 2 players knows and is able to apply the principle of iterated elimination of
dominated strategies. These players will account for the effect of the own choice on
the target number gi and avoid choosing a dominated strategy. But strategic uncer-
tainty concerning the others’ choice questions whether relying on the equilibrium
choice is the best response to what the counterparts who are unaware of the theory
are going to do.
Experimental evidence on guessing games24 all reveals convergence to the equi-
librium, unlike in other games.25 Different explanations for what steers convergence
toward equilibrium have been proposed. It is still debated whether this relies on
22 Cf. Keynes (1936).

23 Cf. Nagel (1995), Duffy and Nagel (1997), Ho, Camerer, and Weigelt (1998) and Weber (2003).
24 See e.g. Camerer, Ho, and Chong (2001), Camerer (2003a, 2003b), and Nagel (1995, 1999).
25 For a review, cf. e.g. Roth (1995).
cognitively capturing iterated elimination of dominated strategies rather than on

non-cognitive behavioural adaptation, as e.g. the application of individual heuris-
tics or mental representation.26
As a possible way of modelling the behavioural patterns that can be observed by
guessing games, Nagel (1995) suggested that individuals adjust their guesses by a
process of naı̈ve best responding rather than by elimination of dominated strategies.
Moreover, Nagel’s model posits that for d = 0 the individuals take 50 instead of 100
as reference point for their reasoning process, whereby each player assumes herself
to be the most sophisticated, i.e. thinks herself to be performing exactly one step of
reasoning ahead of all the others. Nagel’s model fixes 50 as the reference point from
which the individuals start their thinking process, since it assumes level-0 players
to randomly choose a number from the interval [0,100] and therefore, the mean to
equal 50. A level-1 player would then naively best reply, believing everybody else to
be level-0, choosing p · 50. A level-2 player would think one step ahead of a level-
1 player and therefore, guess p2 · 50. In general, a level-k player assumes all the
others to be level-(k-1) and accordingly guesses a number equal p2 · 50 Therefore, a
player who performs infinite iterative steps of reasoning conceptualises and chooses
the Nash equilibrium of 0. The interpretation of the process, which is generated by
iterative best responses as converging toward the equilibrium, relies on the implicit
assumption that different subjects hold different depths of thinking.
By observing the guessing behaviour in different newspaper and lab beauty-
contest experiments27 subjects can actually be clustered at level-1, level-2, level-3,
for then jumping at level-infinity.
Camerer, Ho, and Chong (2001) refuse the idea of behavioural adjustment both
by means of sub-game and of trembling-hand refinements toward the Nash equi-
librium. They alternatively introduce a model of cognitive hierarchy, according to
which a step-k player, perceiving herself to be the most sophisticated, estimates the
other players to be Poisson-distributed from step-0 to step-(k-1).
The parameterization of guessing games has been also proved to influence con-
vergence toward the Nash equilibrium: as Güth, Kocher, and Sutter (2002) pointed
out, “interior equilibria trigger more equilibrium-like behaviour than boundary
equilibria.”28 Because of the individual reluctance to choose the extremes,29 the
heterogeneity of players induces longer deliberation time and higher deviations
from the Nash equilibrium, while continuous payment scheme seems to be a force
which pushes toward equilibrium convergence. Morone and Morone (2007) dis-
cussed Güth et al.’s interior equilibria results in a more general setting and gener-
alised the iterative naı̈ve best replies strategies to a wider class of games, which
showed compatibility among Güth et al.’s and Nagel’s findings on bounded rational
iterative thinking and behaving.
The present experiment exclusively concentrates on first period choices of guess-
ing games with different parameterization, though varying both parameters p and d.
26 Cf. e.g. Camerer (2003a, 2003b) and Sbriglia (2008).
27 Cf. Bosch-Domenech, Montalvo, Nagel, and Satorra (2002).
28 Cf. Güth, Kocher, and Sutter (2002).
29 Cf. e.g. Rubinstein, Tversky, and Heller (1997).
Hereby, because of the modified parameterization and in particular when subjects

are aware of the principle of iterated elimination of dominated strategies, it is fo-
cussed on the changes of behaviour rather than the convergence process. In par-
ticular, how theoretical prescriptions are related and adapted by subjects on their
own as well as others’ behaviour in consideration of the supposed bounded rational
behaviour of the others will be discussed.
7.5.2 Iterated Elimination of Dominated Strategies

in Guessing Games
As a natural first step for exploring theory absorption here symmetric guessing
games will be exclusively focussed on, which are illustrated by:

Γ = N, {Gi }i∈N, {ui }i∈N (7.2)
whereas n is the number of players, gi the set of strategies of player i satisfying

0 ≤ L ≤ H,
gi ∈ Gi = Gi = [L, H] ∀i = 1, . . . , n (7.3)
The guessing game chosen relies on a continuous payoff scheme, according to which
each player
i’s payoff
function depends on her own deviation from the target number
n
g∗ = p ∑
gi
n +d so that
i=1

1 n

u (gi ) = C − c gi − p
∑ i
n i=1
g + d

(7.4)
whereas p ∈ (0, 1), d ≥ 0, L < g∗ < H hold. C is a positive monetary endowment,

and c(> 0) can be interpreted as a fine that subject i has to pay for deviating form
the target number g∗ .
It is most common in guessing games30 that only the player whose guess is clos-
est to the target number g∗ yields a positive payoff. Resembling Güth et al. (2002)
and Morone and Morone (2007) it is, on the contrary, relied on that a continuous
payoff rule, according to which all players win and their deviation from the target
number gets punished.
Starting from the trivial case of being sanctioned for deviating from g∗ if n = 1,
it is easy to state that all numbers g1 belonging to the closed interval [L,H] of the
only player i that are different from the target (gi = g∗ ) are given by:
∗
∗ p d
gi = max min ,H ,L (7.5)
1− p
30 See e.g. Nagel (1995).

For the only player 1 is it easy to recognise g∗1 as the best choice to be made and to
compute it. Therefore, informing the player of the principle of iterated elimination
of dominated strategies leads her to choose g∗1 . The principle of iterated elimination
of dominated strategies can thus be said to be fully absorbable even for boundedly
rational decision makers.
The same principle can be applied also increasing the number of players n ≥ 2.
It prescribes for all players i = 1, . . . , n to eliminate each choice gi fulfilling
p (n − 1) L + nd p (n − 1) H + nd
gi < and gi > (7.6)
n− p n− p
Interestingly, the others are not required to be aware of the principle.
For deriving the unique equilibrium strategies g∗1 = g∗ for all players i =
1, . . . , n, i.e.
p∗d
g∗ = max min ,H ,L (7.7)
1− p
the following assumptions are required:
1. each of the n players i = 1, . . . ,n eliminates dominated strategies repeatedly and
not just once (Principle 1, further P.1);
2. each of the n players i = 1, . . . , n is aware that all other players know that all other
players iteratively eliminate dominated strategies and are aware of all players
doing and knowing that (Principle 2, further P.2).
Thus, even if full absorbability of the iterated elimination of dominated strategies
is, in principle, (i.e. among rational individuals) granted, exploring partial and full
theory absorption of such theoretic principles among bounded rational individuals
is a challenging task. Guessing games qualify as good candidates for this analysis.
If one of the n players is informed about P.1 and P.2, she cannot conclude much
more than that all strategies gi are dominated (what leaves her choice wide open).
On the other hand, informing all of the players about the two principles P.1 and P.2
the immediate choice of g∗ should be expected.
7.5.3 The Experimental Design
The experiment was run in three treatments, each of which contained nine succes-
sive guessing games with different parameters, as illustrated by Table 7.1.
In the table, p and d stand for the parameters that specify each guessing game.
As in Güth et al. (2002), the convergence process is illustrated starting respective
from the interval’s upper (L = 100) and lower (H = 100) bounds. The first rows
refer to what can be eliminated in each successive iteration step-k, while the bottom
row considers step-infinity and therefore, specifies the equilibrium strategy for the
respective game.
Table 7.1 Nine parameterizations of the experimental guessing games

Elimi- pi = 1/2 d = 0 pi = 1/2 d = 25 pi = 1/2 d = 50 pi = 1/3 d = 0 pi = 1/3 d = 25
nation
steps
ki in 0 0.00 100.00 0.00 100.00 0.00 100.00 0.00 100.00 0.00 100.00
1.00 0.00 50.00 12.50 62.50 25.00 75.00 0.00 33.33 8.33 41.67
2.00 0.00 25.00 18.75 43.75 37.50 62.50 0.00 11.11 11.11 22.22
3.00 0.00 12.50 21.88 34.38 43.75 56.25 0.00 3.70 12.04 15.74
4.00 0.00 0.25 23.44 29.69 46.88 53.13 0.00 1.23 12.35 13.58
5.00 0.00 3.13 24.22 27.34 48.44 51.56 0.00 0.41 12.45 12.80
Infinity 0.00 0.00 25.00 25.00 50.00 50.00 0.00 0.00 12.50 12.50
Playing 7 1 4 5 9
order
pi = 1/3 d = 50 pi = 2/3 d = 0 pi = 2/3 d = 25 pi = 2/3 d = 50
ki in 0 0.00 100.00 0.00 100.00 0.00 100.00 0.00 100.00
1.00 16.67 50.00 0.00 66.67 16.67 83.33 33.33 100.00
2.00 22.22 33.33 0.00 44.44 27.78 72.22 55.56 100.00
3.00 24.07 27.78 0.00 29.63 35.19 64.81 70.37 100.00
4.00 24.69 25.93 0.00 19.75 40.12 59.88 80.25 100.00
5.00 24.90 25.31 0.00 13.17 43.42 56.58 86.83 100.00
Infinity 25.00 25.00 0.00 0.00 50.00 50.00 100.00 100.00
Playing 3 2 6 8
order
In each of the tree treatments, every participant played all nine games in the order
specified in Table 7.1 a within subject-design. In each treatment, 32 subjects were
divided into groups of 8, yielding four independent observations. The subjects had
to guess a number from the interval [L,H] whereby as defined above by u(gi), their
earnings were inversely proportional to the deviation of their guess from the target
number, as defined by u(gi) (cf. Eq. 4).
The first treatment (to which it will henceforth referred as UU-treatment)31
served as a control treatment in that all the participants received instructions on rules
of the game but no information concerning the principles of iterated elimination of
dominated strategies.
In the second treatment (the UI- or partial absorption treatment) all subjects were
provided with the instructions to the experiment while only half of them were ad-
ditionally provided with theoretic information regarding the principles of iterated
elimination of dominated strategies. Such information on how to derive the equi-
librium solution was given in the form of tips about the game and also contained a
numerical example for computing the Nash equilibrium.
The tips explained several steps of iterated elimination of dominated strategies
and pointed to the Nash equilibrium, whereas e.g. step-1 was illustrated as follows:
“Any number, chosen by all of the group members won’t exceed 100. This means,
the average won’t exceed 100. The average +d(d = 50) times q (here q=1/3) is 50.
Therefore, the number you should choose, should not exceed 50. If all members of
31 The same labelling of the treatments has been used by Morone et al. (2008).
your group realize this, everybody else will choose 50, and therefore the average will
be 50. Again, your number should not exceed the average +d times q=1/3, which
is 33.33, in order to earn as many points as possible. [. . .]”32 Both “informed” and
“uninformed” subjects (respectively the ones provided with the tips or not) who
were assigned to each group knew how many of the participants were informed and
how many were not. The common knowledge of the dissemination of the theoretic
information intended to put the conditions for testing the principles of iterated elim-
ination of dominated strategies at their partial absorption.
In the third treatment (II- or full absorption treatment) all subjects were informed
on how to iteratively eliminate dominated strategies so as to reach the Nash equi-
librium and knew that all subjects had received the same information. With that, the
third treatment aimed at testing the full absorbability of the guessing game theory.
The participants knew that they would have been randomly matched at the be-
ginning of the experiment and that their group composition would have remained
the same for the whole experiment. Control questions on the basic rules of the game
ensured the understanding of the experimental instructions (but not of the theoretic
information contained in the tips).
After each round, the participants received a feedback on the average of guessed
numbers in their group and their own individual payoff per round to control for
reputation seeking.33 Further, changing parameters were adopted to minimise con-
ditioning from the previous choices.34
The experiment was run in April 2006 at the experimental laboratory of the
Max Planck Institute of Economics in Jena. 96 undergraduate students from Jena
University (32 for each session) took part in the experiment, the computerized exper-
iment has been programmed with the software z-Tree,35 and the ORSEE software36
was used for participants recruitment. With a show-up fee of 4 e the participants
earned in average 8.41 e and the experimental sessions took about 45 min. The
experiment’s instructions are in appendix.
7.5.4 Experimental Results
Using different guessing games, the experiment aimed at testing whether the prin-
ciples of iterated elimination of dominated strategies can be cognitively captured
rather than non-cognitively adapted. As for bounded rational individuals, absorba-
bility of the equilibrium prediction in the form of perfect compliance may not be
granted. The benchmark of perfect strategic interaction has to be compared with the
guesses of bounded rational subjects in the three different treatments.
32 The tips are contained in appendix.

33 Cf. Camerer, Ho, and Chong (2001).
34 Cf. Ho et al. (1998).
35 Cf. Fischbacher (1998).
36 Cf. Greiner (2004).
The concept of thinking depth or of depth of reasoning has been analyzed by

many studies and has been typically applied to the analysis of guessing as well as
to many other games. It refers to the number of steps of elimination of dominated
strategies or, in other words, to the degree of iterated dominance.37 There is much
theoretical and experimental research in exploring the idea of finite depth of rea-
soning, as illustrated e.g. by the theoretical contributes of Binmore (1987), Selten
(1991) and Stahl (1993) and by the experimental studies in Mc Kelvey and Palfrey
(1992), Beard and Beil (1994) or Stahl and Wilson (1994).
However, as the consistency of individual choice behaviour over nine guessing
games with different parameterizations in terms of stability of revealed thinking
depth is unlikely, clustering the subjects according to their depth of reasoning does
not represent a suitable analytic procedure for this experiment. Therefore, the analy-
sis of the experimental data will be restricted by using, instead of the concept of
depth of reasoning, the absolute deviation from the equilibria to catch the level of
application of the principles of iterative eliminating dominated strategies.
Because Nagel (1995) provides evidence for behavioural adjustment toward
equilibrium and ascribes it to qualitative learning which is sensitive to changes in
the parameters, in this spirit, the present study faces the subjects with different one-
period guessing games. Changing parameterizations provides a suitable framing
for testing theory absorption in that it prevents and discourages qualitative learn-
ing and disentangles the effects of revealing the theoretical principles for deriving
the equilibrium.
As follows, the main results on the general effects of providing the individuals
with theoretical information will be discussed with consideration to the payoffs, the
deliberation times and the choices’ deviations from the equilibrium.
7.5.4.1 General Effects of Theory Absorption
Aspects of full theory absorption can be discussed from the comparison between
UU- and II-treatments, while partial theory absorption can be analysed consider-
ing them in respect to the UI-treatment as well as considering the within-subjects
differences in this last treatment.
The experiment provides evidence that revealing the subjects theoretical infor-
mation about the principles of iterated elimination of dominated strategies yields
for (1) smaller equilibria deviations, (2) longer processing time for making a guess,
and (3) higher profits.
The choices’ deviations from equilibrium convey an indicator for appreciating
the absorption of the theory because they can be interpreted as a measure of the
acceptance and compliance with it. For a more robust picture of actual behaviour,
instead of the comparison of the mean of the nine period choices per subject, the
0.25, 0.50 and 0.75 quantile aggregates of the nine period choices per subject will
be focussed on.
37 Cf. Nagel (1995, p. 107).

Table 7.2 First period results of subjects in the three treatments (Morone et al., 2008)
Individual choice Groups’ average Time per subject to Individual profit
deviation from choice deviation choose a guess
the equilibria from the equilibria
Treatment 0.25 0.50 0.75 0.25 0.50 0.75 0.25 0.50 0.75 0.25 0.50 0.75
UU∗ 9.50 16.50 25.00 16.46 18.25 20.48 21.25 32.50 43.50 0.57 21.05 32.85
UI-U 13.25 17.50 25.00 9.84 12.53∗ 12.60 21.50 27.50 47.00 2.34 15.70 28.85
UI-I+ 4.50 12.25 19.75 9.84 12.53 12.60 40.75 57.50’ 71.25 3.55 34.10 40.04
0.00 3.5∗ 15.00 3.66 5.53∗ 8.06 23.75 45.00∗ 63.00 15.27 39.69∗ 44.53
+ + +
II
∗ , ’, and + mark significant differences to either UU (∗ ), UI-U ( ) or UI-I (+ ) subjects
Table 7.2 summarizes the first period results in the different treatments and offers
an overview of the quantile aggregates of individual and groups’ choices deviations
from the Nash equilibrium, the deliberation time per subject and the individual prof-
its. Significant differences (p < 0.05) among treatments are marked as well. The
labels UI-I and UI-U, which are contained in the table and will, for convenience,
be adopted henceforth, respectively refer to the participants to the partial absorption
treatment (UI) who were informed (I) about the theoretic principles for deriving the
equilibrium and those who were uniformed (U).
First period choices capture choice behaviour which was unaffected by learn-
ing and conditioning. Providing theoretic information on how to derive the Nash
equilibrium, a modification in the individual choice behaviour was induced, as (see
Table 7.2) uninformed subjects participating in the UU-treatment deviated signifi-
cantly more from the equilibrium than informed subjects in the absorption treatment
(p < 0.01).38
The same holds for the equilibrium deviations of group averages: it could be ob-
served that the group averages of uninformed (UU) subjects deviated significantly
more from equilibrium than those of informed subjects in the full absorption treat-
ment (II) (p < 0.01). Providing all group members with theoretic information on the
principles of iterated elimination influenced all group members’ behaviour, which
thus reduced the group average equilibrium deviation.
Revealing the principles of iterated elimination of dominated strategies induced
the subjects to think more deeply about the game and about how to choose because a
positive relation between the time needed for making a choice and the thinking effort
put in the task of guessing can be reasonably assumed.39 As shown in Table 7.2,
UU subjects needed significantly less time to type in their guesses than II subjects
(p < 0.01).
After having a closer look at the payoffs, it could be observed that subjects in the
control treatment (II) earned significantly less than subjects in the full absorption
treatment (II) (p < 0.02). This is because first period’s choices of UU subjects ex-
38 Whenever not differently specified, all tests have been performed with an asymptotic Wilcoxon
signed rank test. Further analysis on the experimental data is in Morone et al. (2008).
39 Cf. Güth et al. (2002).
hibited a significantly larger variance of the discrepancy of individual choices than

groups’ averages (p < 0.002). Specifically, the absolute variance of UU subjects’
choices was 286.74, while the one of II subjects was 85.84. This reveals the imme-
diate impact of providing individuals with theoretical information, which affected
the individual choice behaviour from the first period on. Hence, this could be inter-
preted as a sign of propensity to accept and apply the theory, trusting the others to
accept it and apply its prescriptions, as well.
Informing the subjects about the principles of iterated elimination of dominated
strategies induced an immediate change in their choice behaviour. In homogenously
informed groups (in the II treatment), the subjects self-refer the game theoretical
predictions on themselves and suppose the others to do the same. The subjects pro-
jected their own choice onto the choice of similarly informed group members. The
theoretic principles on how to derive the equilibrium worked in a stabilizing way
on the beliefs about the others’ choices and increased the trust in the possibility of
predicting them with a certain degree of accuracy40 possibly because a commonly
known pattern of reasoning (embodied by P.1 and P.2) had been provided. The ad-
vantage of being informed can be ascribed to the regular application and trust in the
others to apply the theoretical prescriptions. As informed subjects expected more
choices to be close to the equilibrium, the groups’ variance of choices decreased
leading, therefore, to higher payoffs.
These considerations of the effects of informing the subjects about the guess-
ing game theory hold even in respect to the choices in all periods. As shown by
Table 7.3, the choice behaviour over all periods seems to confirm the tendencies
revealed by the first period results.
In all of the successive periods, the subjects in the control treatment (UU) re-
vealed a higher choice deviation from the equilibria than II subjects (p < 0.01).
Groups’ averages deviated from equilibrium significantly more in the UU than in
the II treatment (p < 0.001), too. In all periods the deliberation time was signif-
icantly lower among uninformed subjects participating in the UU treatment than
among informed subjects in the full absorption treatment (p < 0.01). Overall profits
of II subjects were higher than those of UU subjects (p < 0.01).
Table 7.3 All period results of subjects in the three treatments (Morone et al., 2008)
Individual choice Groups’ average Time per subject to Individual profit
deviation from the choice deviation choose a guess
equilibria from the equilibria
Treatment 0.25 0.5 0.75 0.25 0.5 0.75 0.25 0.5 0.75 0.25 0.5 0.75
UU∗ 8.38 19.9 33.00 11.94 18.50 26.91 16.75 24.00 36.00 0.00 24.57 38.20
UI-U 8.00 17.00∗ 27.00 10.12 14.58∗ 20.91 20.00 32.00∗ 52.25 4.83 29.02∗ 39.33
UI-U+ 5.00 14.75’ 25.00 10.12 14.58 20.91 26.00 43.50’ 60.00 9.06 26.86 38.82
10.00∗ 10.96∗ 43.00∗ 30.84∗
+ + + +
II 0.00 25.00 7.42 18.20 25.00 64.00 9.17 42.71
Note that ∗ , ’, and + , mark significant differences (p < 0.05) resp. to UU (∗ ), UI-U (’) or UI-I (+ )
subjects
40 Cf. Dawes (1988).
7.5.4.2 Equilibrium Behaviour
Focussing on equilibrium choices has been interpreted as a way of disentangling

learning effects from immediate theory absorption, i.e. the immediate jump to the
equilibrium choice. In this sense providing theoretic information associates overall
with an increasing number of equilibrium choices.
As illustrated by Table 7.4, in the first period the subjects in the full absorption
treatment (II) chose the equilibrium more often than the subjects both in the partial
absorption treatment (UI) and in the control treatment (UU). Both differences are
significant with p < 0.01. The subjects in all-informed groups performed signifi-
cantly better than the uniformed subjects (p = 0.011).
In the UI-treatment the number of the first-period equilibrium hits did not differ
significantly from their number in the UU-treatment. However, this was revealed to
be quite high possibly because of the high number of economics and natural science
students among the participants.
A binomial test both on the first and on all successive periods revealed no
constant equilibrium choice behaviour of any subject in any of the treatments
(p < 0.001). Hence, the principles of iterated elimination of dominated strategies
cannot be claimed of having been fully absorbed by the bounded rational experi-
mental subjects. Full absorbability would have required the individuals to comply
with the theory’s advice. In a satisficing perspective, the compliance with the theory
can be translated in the elimination of strategies until further elimination won’t pro-
duce higher payoffs. At least when all players are known to be informed about the
theoretic principles for inferring the Nash equilibrium, it could have been reason-
able, even in a satisficing approach, to expect the individuals to jump immediately
to the equilibrium from the very first period on.
Still, the significant differences between the subjects’ behaviour among treat-
ments prove that the individuals actually perceive the reflexive character of the
game’s theoretic propositions, which are related to the context faced and induce a
modification of behaviour. The individuals modify their behaviour in consideration
of the others’ expected choices. The others’ behaviour is anticipated on the basis of
the individual assumptions on the other subjects’ limited absorption capabilities and
of the doubts on their acceptance of the theoretical advice.
In the experiment the modifications in the individuals’ choice behaviour (il-
lustrated e.g. by equilibrium choice deviation) were inversely proportional to the
disseminated knowledge: while the average first period choice deviation from equi-
librium was 5.53 among subjects in the II treatment, it was 9.97 among UI and 18.7
among UU.
Table 7.4 Percentages of equilibrium choices in the three treatments (Morone et al., 2008)
Period 1 2 3 4 5 6 7 8 9 Total
UU∗ 15.63 6.25 6.25 34.38 3.13 21.58 12.50 0.00 0.00 11.11
UI 12.50 3.13 28.13 9.38 6.25 9.38 6.25 6.25 3.13 9.38
II 43.75∗ ’ 6.25 59.38 40.63 6.25 28.13 9.38 21.88 12.50 25.35∗ ’
∗ and ’ mark significant differences to either UU (∗ ) or UI (’) subjects
Analysing choice behaviour over all periods corroborates similar conclusions, as

in the full absorption treatment the subjects opted for the equilibrium significantly
more often than in the partial absorption treatment (p < 0.001) and in the control
one (p = 0.003).
The common knowledge of the heterogeneous information dissemination
(i.e. having informed all subjects in the UI treatment that half of them received
theoretic information and half did not) significantly reduced the overall quote of
equilibrium hits also in comparison to their number in the UU treatment (p < 0.001).
The frequency of equilibrium choices in the UI treatment remains however higher
among informed (henceforth labelled as UI-I) than among uninformed (UI-U)
subjects (p < 0.001), adding respective to 21 and 6 equilibrium hits.
Thirty-eight of the 432 choices (8.79%) made by the 48 uninformed subjects
equal the equilibrium, while 47.92% of the uninformed subjects never hit the equi-
librium. During the nine periods, the 48 informed subjects chose the equilibrium 94
times, i.e. by the 21.76% of cases, while only 33.33% of them never typed in an
equilibrium guess.
In consistency with the findings of previous studies,41 interior equilibria were
chosen significantly more often than border equilibria (p < 0.001). This provides
further evidence for interior solution to be easier to calculate than solutions which
lie on the boundaries. The relative frequencies of equilibrium choices in border and
interior equilibria games in the different treatments are represented in Fig. 7.2.
The differences in the equilibrium choice behaviour reveal that providing the
respondents with theoretical information induces a significant increase of equilib-
rium choices (p < 0.001). Instant changes in the choice behaviour acknowledge the
Relative frequency of Nash equilibrium hits

0.3
0.25
relative frequencies
0.2
0.15
0.1
0.05
0
border interior total border interior total border interior total
UU session UI session II session
Fig. 7.2 Relative frequency of equilibrium hits in games with interior and border solutions
Morone et al. (2008)
41 Cf. Güth et al. (2002) and Ho et al. (1998).

cognitive capturing of the principles of iterated elimination of dominated strategies,

as revealed by the hypothesis of equivalent equilibrium choice behaviour between
the II- and the UU-treatment which could be rejected on the 1% level for the first
period and on the 5% level over all nine periods.
7.5.4.3 Partial Theory Absorption
The analysis of partial absorption of the guessing game theory is revealed to be a

particularly challenging task, since heterogeneous distribution of the theoretic in-
formation can raise the strategic uncertainty regarding the behaviour of the others.
The experiment gives evidence first for the definite knowledge about the hetero-
geneous information structure of the group to induce informed subjects (UI-I), to
deviate more from equilibrium than the uniformed subjects (UI-U). Second, it ac-
counts for similar time consumption by both kinds of subjects, which was higher
than in the control treatment. Third, the profits in the UI-treatment were similar to
those in the UU-treatment and lower than in the II-treatment.
Starting by observing the first period choice behaviour (see Table 7.2), it can
be noted that informed UI subjects’ choices deviated more from equilibrium than
those of the informed II subjects (p < 0.01), while the choices of the uninformed
respondents in the UI-treatment were similar to those in the UU-treatment. A further
interesting feature of the UI-treatment is that both informed and uniformed subjects
acted similarly. Being aware of the presence of subjects who were not instructed
on how to derive the equilibrium solution, the informed UI subjects’ predictions
contemplated quite large equilibrium deviation of the uninformed subjects. The in-
formed subjects adapted their choice to this expectation in the hope of sustaining
profits. The uninformed subjects in the UI-treatment acted similarly to the subjects
in the UU-treatment and needed the same time to make a guess, while their informed
counterparts took significantly more time to deliberate a choice (p < 0.023).
Group averages’ deviations from equilibrium were significantly smaller in
the UI- than in the UU-treatment because informed UI-subjects typed slightly
lower numbers than uninformed UI-subjects (this difference being however non-
significant). Consequently, the informed subjects in the UI-treatment earned signif-
icantly less than the informed subjects in the II-treatment (p < 0.03).
In summarization, no significant differences could be observed between the first
period choices of the uninformed subjects in the UI- and in the UU-treatment, or be-
tween those of the uninformed and informed subjects in the UI-treatment. This indi-
cates that subjects possessing superior theoretic information tried, when interacting
with less informed individuals, to anticipate their mental limitations and correspon-
dently modified their choice behaviour, but could not avoid profit losses compared
to superiorly informed subjects in homogeneous groups.
Considering the choice behaviour of the uninformed UI-subjects during all pe-
riods (as shown by Table 7.3), significantly smaller deviations from the equilib-
rium than among the uniformed UU-subjects can be noted (p < 0.02). This can
be ascribed to the awareness of having to interact with superiorly informed sub-
jects and could be derived in absence of a theory to be captured from non-cognitive

idea adaptation.42 Because evidence for this effect can be only found if consider-
ing the overall period choices, such a non-cognitive adaptation seems to be based
on an improvement of the individual heuristics or changes in the mental represen-
tations of outcomes. Overall, however, partial theory absorption lowered the profits
which increased the likelihood of deviation from equilibrium because the informed
UI-individuals did not reasonably expect their uniformed counterparts to be able
to infer the iterative principles of eliminating dominated strategies. Informed UI-
subjects adapted their choice behaviour to such expectations so that their choices
deviated significantly more from the equilibrium than choices of the informed II-
subjects (p < 0.05). Therefore, informed subjects in the UI-treatment can be con-
sidered to steer the higher group averages in the UI-in comparison to the II-treatment
(p < 0.012). Still, the average absolute deviations from the equilibrium were lower
than those of the groups belonging to the control treatment (p < 0.01).
Interestingly, uninformed UI-subjects took more time typing in their choices than
UU-subjects (p < 0.01). This indicates that knowledge about the heterogeneous
distribution of theoretical information influenced the deliberation time taken by the
uninformed subjects, which challenged them to think harder about the game and
how to play. While the informed subjects in the UI-treatment needed more time
than their uniformed group members (p < 0.01), they required less time than the
informed II-subjects (p < 0.05).
In regard to the profits, they were similar within the UI-treatment, whereas the
uninformed subjects gained more in the UI- rather than in the UU-treatment (p <
0.05), while the informed UI-subjects yielded significantly lower profits than what
II-subjects (p < 0.05).
The informed subjects who were assigned to the partial absorption treatment car-
ried the burden of their groups’ uninformed members in the form of profit losses.
While the uninformed subjects in heterogeneous groups adapted their choice behav-
iour without lowering (compared to the uninformed subjects in UU-groups) their
earnings, the informed subjects in UI-groups suffered profit losses compared to
the subjects interacting in II-groups. This can be alleged to the high strategic un-
certainty in the heterogeneous groups regarding the uninformed subjects’ behav-
iour. Higher time consumption and smaller equilibrium deviation of uninformed
UI-subjects compared to UU-subjects can be claimed to depend on an essentially
non-cognitive behavioural adaption.
7.5.5 Conclusions
The game theoretical predictions for the guessing game were used in this exper-
iment to test them at their absorbability and inform only a part or the totality of
the individuals interacting. For disentangling the effects of theory absorption from
42 “With no theory to capture, this ought to be a consequence of non-cognitive idea adaptation.”
Cf. Morone et al. (2008, p. 172).

those of learning the experiment comprised nine one-period guessing games with
different parameterizations.
As a benchmark for tackling this research aim, a control treatment was run show-
ing that by solely providing the basic rule of the guessing game, a more or less ran-
dom choice behaviour emerged in a similar setting. On the contrary, in the groups
all of whose members were provided with theoretical information on how to derive
the guessing games’ equilibrium solutions, both the individual guesses and groups’
averages were significantly closer to the equilibrium. This could be observed in the
very first as well as in all the other periods and hints at the cognitive capturing
of the theoretic principles of iterated elimination of dominated strategies and the
self-reflection of their advice. Although a significant impact of revealing theoretic
information to all group members was observed (in terms of increased frequency
of equilibrium hits, longer processing times, lower group average deviation from
the equilibrium and higher profits compared to the results of the control treatment),
the hypothesis of full theory absorption, in the sense of perfect compliance to the
theory’s prescriptive, could not be acknowledged.
It can be assumed that the subjects interacting in all-informed groups did not
choose to perfectly adhere to the theory in the hope of satisfying their profit as-
pirations. Assuming limited capabilities in the other group members, the informed
subjects adapted their choice behaviour, in this way revealing the cognitive captur-
ing of the theoretic principles with adaptation of their predictions to the capabilities
ascribed to the others.
Being aware of the partial dissemination of theoretic information raised strategic
uncertainty and diminished trust in the uninformed group members’ capabilities.
Assuming the same cognitive capturing of the theory by the subjects in the full and
in the partial absorption treatments, the informed subjects in the latter adapted the
theoretic prescriptions even more because of the presence of uninformed subjects
in order to sustain their profits. Being aware of interacting with superiorly informed
counterparts induced the uninformed UI-subjects to modify their choice behaviour
in an essentially non-cognitive way. In comparison to the uninformed subjects in
the control treatment they needed more time to type guesses which deviated less
from the equilibrium in all periods and therefore, steered group averages toward the
equilibrium.
Therefore, partial theory absorption can be linked to a combination of non-
cognitive ideas adaptation by uninformed subjects and cognitive alternations of the
captured principles by informed subjects, whereby the degree to which a theory’s
predictions are adapted depends on the commonly known information dissemina-
tion. The trust in the others’ capabilities of capturing the principles of iterated elim-
ination of dominated strategies was obviously lower in case of partial rather than full
information dissemination. This, adding to the consideration of the bounded ratio-
nality of the others, led to revisions of the game’s theoretic principles and motivated
the observed violation of perfect theory absorption.
The three main findings from this experiment can be summarized as follows:
1. Subjects who received theoretical information about the principles of iterated
elimination of dominated strategies revealed smaller deviations from equilibrium
7.6 Absorbability of Herd Behaviour and Informational Cascades Theories 171
choices, needed longer processing time, and higher profits in the first and in all
periods.
2. Theoretic information triggers a higher number of equilibrium choices.
3. “Definite knowledge about the heterogeneous information structure of the group
induces (un-)informed UI-subject to exhibit (lower) higher choice deviations
from equilibrium, (higher) similar time consumption, and (similar) lower profits
than (UU) II subjects.”43
7.6 An Experimental Study on the Absorbability of Herd

Behaviour and Informational Cascades Theories44
This experiment discusses the absorbability of informational cascades’ theory by

bounded rational decision-makers. In this insight, it analyses whether providing in-
dividuals with theoretic information on informational cascades affects the overall
probability of the herding phenomena to occur as well as whether an incorrect cas-
cade can be reversed because of bounded rational adapting of the theory’s prescrip-
tive. In particular, the occurrence of informational cascades in an experimental in-
vestment task will be observed. The conception which underlies this analysis is that
providing the subjects with theoretic information on probability assessment could
make them aware of the fragility and idiosyncrasy of informational cascades and
thus affect the probability of (erroneous) cascades to occur.
As follows, after a short review of studies on herding and informational cascades,
a simple model will provide the theoretical framework for explaining informational
cascades. An overview of related experimental evidence will be presented, followed
by the design and the main results of this experiment.
7.6.1 Herding and Informational Cascades
Conformity and fluctuations in mass behaviour are frequent features which char-
acterize many social and economic situations.45 Individuals are influenced by the
behaviour of the others, as it can be informative to many extents and promote what
has been depicted as “social learning.”46 Trying to learn from the others typically
induces imitative behaviour which can under some circumstances be rational even
when it implies choosing differently rather than solely relying on one’s own infor-
mation. Despite individual rational behaviour, erroneous, inefficient outcomes can
arise on this basis.
43 Cf. Morone et al. (2008, p. 171).

44 This experiment and its results are discussed in Fiore, Morone, and Sandri (2007).
45 Cf. e.g, Avery and Zemsky (1998), Çelen and Kariv (2004), Scharfstein and Stein (1990) and
Welch (2000).
46 For surveys, see e.g. Douglas (1996) and Bikhchandani et al. (1998).
The phenomenon of “herding” has been widely analysed in the last decade. The
notion of “herd behaviour” refers to the phenomenon according to which people
ignore their own private information in order to follow the example of others. Becker
(1991) first pointed out this kind of behaviour, whose analysis was then further
developed by Banerjee (1992) and Bikhchandani, Hirshleifer, and Welch (1992).47
Herding model and informational cascades are interactive, have a clear economic
interpretation, are very simple to explain and represent a setting in which the ratio-
nal action is independent by the subjects’ preferences. These features make them
particularly attractive for studying rationality and learning, in general, and interest-
ing for experimentally testing theory absorption among bounded rational decision
makers.
In addition, as herd behaviour applies to a conspicuous number of social and
economic issues, investigating how it can be influenced by theory absorption could
yield useful implications, even for defining procedures for preventing erroneous
cascades to occur or at least for containing their frequency or dimensioning their
consequences.
In many real-life situations, people have to make their decisions sequentially.
They can observe the decision taken by previous subjects and often be influenced
by that. For example, when having to choose between two restaurants, in absence
of other information, people are more prone to prefer the more crowded one. This
behaviour is motivated by the assumption, which intuitively sounds rational, that the
number of diners reflects the quality of the restaurant. As this kind of logic inspires
all subjects, if a few subjects randomly decide to enter the restaurant, then all later
subjects will join the queue, which starts a perverse dynamics according to which
every new subject joining the queue raises the probability for others to queue up.48
Ignoring private information for joining the queue is the basic mechanism of herd
behaviour. Assuming rational expectations implies that an agent could compensate
her lack of knowledge on the true model of the situation she is confronted with, and
thus reach an efficient outcome by drawing on all available information, therefore,
encompassing the observation of the others’ choices.
In a setting in which individuals act sequentially and have common knowledge
of the history of the game,49 the players will inform their decision-making both to
their private information and to the actions of their predecessors, without knowing
which information the choices were based upon. As a result, the choices of the
individuals will be correlated, even if their personal information and background
differ, and mistakes by early decision-makers might be transmitted to latter ones,
delaying in all likelihood the settlement for an efficient outcome or even preventing
it altogether.
Two simple frameworks for modelling herd behaviour are proposed by Banerjee
(1992) and Bikhchandani et al. (1992). They both refer to a setting in which sequen-
47 Examples of experimental investigations of herding behaviour are among others Allsopp and
Hey (2000), Anderson (2001), Cipriani and Guarino (2005), Huck and Oechssler (2000), Hung
and Dominitz (2004), Stiehler (2003).
48 Cf. Becker (1991).
49 By common we mean that at time t player t knows all the actions taken by previous players.
tially acting agents have to choose a winning action a among a set of alternatives.
These two models differ in that while Bikhchandani et al. assumes the wining ac-
tion a to be an element of the interval [0, 1] ∈ N and all subjects to receive a signal,
in Banarjee’s model a belongs to the set [0, 1] ∈ R. The individuals can be either
informed or uninformed, i.e. they can either have received a signal or not.
Further, Bikhchandani et al. (1992) explore the concept of informational cas-
cades. In the process they try to explain not only conformity among agents but also
“rapid and short-lived fluctuations such as fads, fashions, booms and crashes.”50
They point out that the conformity of followers in a cascade contains no informa-
tional value. In this sense, the cascade is fragile because it can be upset by the arrival
of new public information (while it cannot be reversed if superior information is not
provided). It is also idiosyncratic “in that random events combined with the choices
of the first few players determine the type of behaviour on which individuals herd.”51
Based upon the model of Bikhchandani et al. (1992), Willinger and Ziegelmeyer
(1998) developed a framework with asymmetric accuracy of private information in
order to experimentally investigate the aspect of fragility of informational cascades.
Their analysis revels that providing the subjects who decide immediately after the
occurrence of a cascade with an additional signal lowered the occurrence of infor-
mational cascades and interrupted herding.
7.6.2 A Simple Model of Informational Cascades: A Dichotomy

Choice Model
An informational cascade, which occurs when people prefer to follow the behav-
iour of the others ignoring their own information, can be modelled by the follow-
ing game.
Consider N players who have to choose sequentially among two options, namely
urn B (for black) and urn W (for white). While urn B contains two black and one
white balls, urn W has two white balls and a black one. Player 1 draws a ball from a
randomly chosen urn and has then to guess from which urn she has pulled the ball.
A correct guess will yield her a payoff of 1, a wrong guess will yield a 0. Player
2 who can observe player 1’s guess has to pull a ball from the same urn and then
to guess. Player 3 observes the choices of both previous players, draws a ball and
makes her choice and so on until player N who can type in her guess by having
observed the choice of all (N-1) precedent players.
As it is rational of player 1 to guess the black (white) urn when she draws a black
(white) ball, player 2 is therefore confronted with one of the following scenarios:
1. she observes player 1 having chosen the black urn, and draws a black ball;
2. she observes player 1 having chosen the black urn, and draws a white ball;
50 Cf. Bikhchandani, Hirshleifer, and Welch (1992, p. 994).

51 Cf. Bikhchandani and Sharma (2001, p. 284).
3. she observes player 1 having chosen the white urn, and draws a black ball;
4. she observes player 1 having chosen the white urn, and draws a white ball.
It is therefore rational for player 2 to choose the black urn in the first scenario and
to opt for the white urn in the fourth scenario. On the contrary, in the second and
third scenarios she should be indifferent to the two urns and choose among them
with equal probability.
Under these assumptions and after two individuals have played, the unconditional
ex-ante probabilities of an informational cascade choosing the white urn (“White-
cascade”), the black urn (“Black-cascade”), as well as the probability of no cascade
to occur (“No-cascade”) are calculated by Bikhchandani et al. (1992) as follows:
1−p+p2
White − cascasde = 2 ; No − cascasde = p − p2 ;
2
Black − cascade = 1−p+p
2 (7.8)
After an even number of players (n = 2 m) have played, they calculate these proba-
bilities to become:
1−(p−p2 )m
White − cascasde = 2 ; No − cascasde = (p − p2 )m ;
2 )m
Black − cascade = 1−(p−p 2 (7.9)
where p stands for the probability of observing a correct signal. As a result, the
bigger p is, i.e. the more accurate the signal is, the sooner an informational cascade
can start (see Fig. 7.3).
1,9998
1,9996
1,9994
1,9992
Fig. 7.3 Probability of start-

ing a cascade as a function of
p, the correctness of the signal 0.5 0.6 0.7 0.8 0.9 1
(N = 10) n
Further, Bikhchandani, Hirshleifer and Welch calculate the probability of ending

up in the correct cascade after two players have chosen, given that the chosen urn is
the white one. These are respectively the calculations:
p(p+1)
White − cascasde = 2 ; No − cascasde = p(1 − p);
Black − cascade = (p−2)(p−1)
2 (7.10)
In the general case, i.e. after an even number of players (n = 2 m), these probabilities
can be generalized as:
p(p+1)[1−(p−p2 )m ]
White − cascade = 2(1−p+p2 )
(7.11)
No − cascade = (p − p2 )m (7.12)
(p−2)(p−1)[ 1−(p−p2 )m ]
Black-cascade = 2(1−p+p2 )
(7.13)
Although the probability of observing a correct cascade, illustrated by Eq. 11, in-
creases in p and m, even for very informative signals (i.e. p close to 1), the proba-
bility of a wrong cascade remains remarkably high, as shown by Eq. 13.
Figure 7.4 represents the probability of a correct and incorrect cascade for
N = 10.
0.8
0.6
++
++
++
0.4 ++
++
++
++
++
++
++
+++
0.2 ++
+++
+++
+++
+++
++++
++++
0 +++
0.5 0.6 0.7 0.8 0.9 1
P
Fig. 7.4 Probability of a correct (continuous line) and incorrect cascade (dotted line) as a function
of p, the correctness of signal (N = 10)
7.6.3 Experimental Design
The experiment included two treatments: while the control treatment is based on
the model in Bikhchandani et al. (1992), in the absorption treatment the participants
received theoretical information about informational cascades. More precisely, in
the absorption treatment the respondents were provided with an illustration of how
to infer the expected value of adoption and rejection in dependency of the accuracy
of the private signal and how to deduce the individual optimal decision rule. The
absorption treatment informed the experimental hypothesis of the theoretic infor-
mation provided to prevent incorrect cascades to occur.
The experiment was programmed using the software Z-tree52 and was run in
March 2007 at the laboratory of ESSE at the University of Bari.
Each treatment took about an hour and was made up of 22 periods. Since the first
two periods were trials, the final payment was made at the end of each treatment only
on the 20 real game-playing periods. The experiment’s instructions are contained in
appendix.
N = 10 subjects participated in each session: each of them sat in front of a PC
connected by a net and could neither see the others nor communicate with them. The
participants were all undergraduate students in economics, who were not familiar
with similar experiments.
In each period, which lasted for about two minutes, the subjects sequentially
played in a randomly determined order. A message on their PC screen informed
them when it was their turn. The subjects were asked to decide whether to invest
in a new product or not, not knowing whether the new product would have been
profitable or not. They knew that the profitability of the product was dependent on
two equally likely events and that if the product would have been successful (V = 1),
each player would have gained 0.5 e in case of having invested, and zero otherwise.
The opposite would have been true if the product would not have been successful
(V = 0), each player would have gained 0.5 e in case of not having invested (which
would have been the right decision), and zero otherwise. In order to exclude losses
by the participants, no cost of adopting was considered. The true value of V was
exogenously determined in each period but was not revealed to the subjects, who
only saw a free-of-charge signal S and knew it had a probability p of 0.75 of being
correct.
In each period of the control treatment, the subjects were informed about their
own turn to play, all previous guesses, and their own signal S.
In addition to such information, the subjects in the absorption treatment received
a decisional aid in the form of tips about the game. It contained theoretical infor-
mation on how to derive, in dependency of the individual position in the queue, the
unconditional ex-ante probabilities of ending up respectively in a correct or in an
incorrect cascade.
52 Fischbacher (1998).
In both treatments the subjects were informed at the end of each period about the
true value of V and their individual period-payoff. After all periods were played, the
subjects were paid and could leave the laboratory. Average earnings were 7 e.
7.6.4 Results
The experimental results enable to discuss some aspects of absorbability of infor-

mational cascades’ theory by bounded rational decision-makers. The experimental
hypothesis which argues whether providing individuals with theoretic information
on informational cascades affects the overall probability of herding phenomena and
that of their reversal will be in particular discussed focussing on the general effects
of theory absorption, comparing the social efficiency of outcomes among treatments
and testing for directional learning. The main results, this experiment accounts for,
can be discussed as follows.
7.6.4.1 General Effects of Theory Absorption
Considering the benchmark provided by the theory on herding (and informational

cascades), it is possible to qualify the individual choices observed in the experiment
as rational or irrational, respectively if they are conform to the optimal strategy or
not. In the experimental simple set-up, where the two states of nature are equally
probable and the private signals identically distributed, the optimal strategy in a
Bayesian sense can be defined taking into account the decision of the predecessors
and the individual’s private information, as doing the count on the previous deci-
sions (the one’s own signal being included)53 and adopting the most chosen option.
The adoption of the tie-breaking rule if indifferent has been assessed as rational
(optimal) behaviour, both if generating a cascade or not. Further, the case in which
“an imbalance of previous inferred signals causes a person’s optimal decision to be
inconsistent with his or her private signal”54 has been considered as a cascade.
Choice which were not consistent with these rules has been qualified as irrational,
whereas it has been distinguished between two subspecies of irrational behaviour,
which have been labelled as “signal-keeping” and “not-rationalized.” They respec-
tive correspond to the cases in which following one’s own private signal can provide
a somehow logical explanation of the individual’s choice and to those in which there
is no plausible explanation for it.
According to these criteria the choices of the experimental subjects can be
grouped and summarized as shown by Table 7.5.
It can be clearly noted that providing the respondents with theoretical information
about the optimal Bayesian strategy yields for higher consistency of behaviour with
53 Cf. Anderson and Holt (1997).

54 Cf. Anderson and Holt (1997, p. 851).
Table 7.5 Classification of the individual behaviours observed

Signal Rational (Bayes’ rule) Irrational
Not rationalized Signal-keeping
Control Treatment 0,72 0,725 0,15 0,135
Absorption Treatment 0,715 0,935 0,035 0,04
Table 7.6 Perceptual occurrence of informational cascades

Cascades occurrence Correct cascades Wrong cascades
(%) (%) (%)
Control Treatment 48.07 24 76
Absorption Treatment 86.66 49.02 50.98
the theoretical predictions and, among the irrational choices, for the lowering of
adoption of non-rationalized behaviours.
The higher compliance with the Bayesian optimal strategy provides evidence for
the absorption of the theoretical predictions55 .
A further interesting feature which emerges from the experimental data is that
while the frequency of optimal behaviour is higher in the absorption than in the con-
trol treatment, the overall frequency of signal following is almost the same among
treatments (72% in the control treatment versus 71.5% in the absorption one). There-
fore, the theoretic information did not affect the propensity of the individuals to rely
on their own private information.
The two treatments gave account for a different occurrence of informational cas-
cades. It has been in particular considered, how many informational cascades which
could have formed occurred in fact. In this insight, a cascade has been considered as
possible to occur whenever the choice between following one’s own private infor-
mation and Bayesian optimization are mutually exclusive. In all of these cases, an
informational cascade takes place if the individual ignore her own signal and prefer
to herd.
According to these criteria, while in the control treatment 25 of the 52 possible
informational cascades formed, in the absorption treatment 51 out of the 59 infor-
mational cascades that would have been possible established. This acknowledges
for the percentages which are shown by Table 7.6.
Fragility of cascades is higher in the control treatment, in which in particular it
never happened that a cascade starts at the beginning of the period and last until the
end of it. Instead, in the absorption treatment, informational cascades and herding
seem to be more difficult to reverse. In this insight, it should be however considered
that the occurrence of correct cascades is higher in the absorption than in the control
treatment: while in the control treatment only 9 of the 25 informational cascades
which affirmed were correct, in the absorption 25 correct and 26 wrong cascades
formed. Table 7.6 summarizes the relative occurrence of correct and wrong cascades
in perceptual values.
55 At a 0.01 significance level.
7.6.4.2 Social Efficiency
For testing if proving the respondents of theoretical information on the optimal

decision-rule enhances their profits, improves their winning chances and better
therefore the efficiency degree of the outcome, the earnings distribution per posi-
tion in the queue per treatment can be compared (cf. Fig. 7.5).
We can accept at a 95% significance level that earnings are higher, on average, in
the absorption than in the control treatment.
The percentage of winning per position in the queue (cf. Fig. 7.6) which can be
considered as a proxy for the individual utility is clearly higher in the absorption
than in the control treatment, this difference being significant with p < 0.043.
In this sense, providing the individuals with theoretical information on herding
behaviour and informational cascades reveals to be a device which can be preferable
both from the social and from the individual point of view.
10
9
8
avarage earning
7
6
Control Treatment
5
Absorption treatment
4
3
2
1
0
1 2 3 4 5 6 7 8 9 10
Position in the queue
Fig. 7.5 Average earnings per position in the queue per treatment
1
0,9
winning percentage
0,8
0,7
Control treatment
0,6
0,5 Absorption
0,4 treatment
0,3
0,2
0,1
0
1 2 3 4 5 6 7 8 9 10
position in the queue
Fig. 7.6 Percentages of winning per position in the queue per treatment
1
Control
0,9
Treatment
optimal strategy (%)
0,8
0,7 Absorption
Treatment
0,6
0,5
0,4
0,3
0,2
0,1
0
1 2 3 4 5 6 7 8 9 1011121314151617181920
period
Fig. 7.7 Percentages of optimal strategy per treatment
0,4
deviation from the optimal
0,35
0,3
0,25
strategy
Control Treatment
0,2
Absorption Treatment
0,15
0,1
0,05
0
[1 - 5] [6 -10] [11-15] [16- 20]
classes of periods
Fig. 7.8 Average percentages of deviant strategies per classes of periods
Figure 7.7 illustrates the percentages with which the winning strategy has been
per period adopted in each treatment. At a significance level of 99% higher adoption
of the optimal decision-rule implies that the individuals are able to conceptualize the
theory and to refer it on the setting they face.
7.6.4.3 Learning
In order to appreciate whether learning effects occurred in the experimental setting

the 20 game-playing periods can be divided onto 4 groups. The average percentages
of deviations from the optimal strategy per classes of periods (see Fig. 7.8) does
seem to corroborate the idea of a marked learning process, neither in the control nor
in the absorption treatment. In particular in the control treatment, a slight tendency
toward the reduction of deviant choices can be noted.
7.6.5 Conclusions
Herding behaviour and informational cascades refer to cases in which individual

rational behaviour may result in a non-optimal strategy at aggregate level. As sug-
gested by Becker (1991), these phenomena deal with situations in which information
can be somehow linked with negative externalities. By conforming to the behaviour
of the preceding subjects ignoring ones’ own private information the behaviour of
the others stops up to a certain point to be informative. Some studies on herding ar-
gue if “society may actually be better off by constraining some of the people to use
only their own information.”56 This has been also acknowledged by some empirical
evidence.57 Shifting the perspective, the present study investigates whether provid-
ing the individuals acting in settings in which herding and informational cascades
are likely to occur with the theoretical principle of Bayesian optimization constitute
an alternative mechanism for improving the degree of social efficiency.
The evidence from this experiment seems to corroborate this idea, in that average
earnings and the percentage of winning per position in the queue were higher among
the responders who received the theoretic information than among those who did
not. In this sense, providing the individuals with theoretical information on herding
behaviour and informational cascades reveals to be preferable both from the social
and from the individual point of view.
The experimental evidence further proves that theoretic information on infor-
mational cascades affects the overall probability of herding phenomena and that of
their reversal. It could be in particular observed that the behaviour of the respondents
who received the theoretic information revealed a higher consistency and compli-
ance with the theoretical prescriptions and yields for lowering the occurrence of
non-rationalized behaviour. The higher compliance with the Bayesian optimal strat-
egy corroborates the hypothesis of absorbability of informational cascades’ theory,
revealing its understanding and acceptance by bounded rational individuals. The
theoretic information did not affect the propensity of the individuals to rely on their
own private information.
The overall occurrence of informational cascades was higher among informed
than among uniformed individuals, whereas the frequency of correct cascades was
more than double. Thus providing the individuals with theoretical information on
herding behaviour and informational cascades reveals to be a device which improves
the occurrence of cascades that can be associated with individually and collectively
favourable outcomes. Informational cascades seem however to be less fragile and
more difficult to be reversed among individuals who are aware of the theoretical
prescriptions.
The experimental results provide, in none of the treatments, significant evidence
for marked patterns of directional learning concerning the application of the optimal
decision-rule.
56 Bannerjee (1992, p. 798).

57 Cf. e.g. Fiore and Morone (2007).
7.7 Concluding Remarks on the Experimental Examination
What emerged from the experimental examination of the self-referentiality of eco-

nomic theories and their absorption confirms so far the intuition that the investi-
gation should include a large spectrum of experiments. The experimental evidence
corroborates the thesis that absorbability of economic theories is, in principle, al-
ways possible and that its concrete modality are context specific and depends on the
complex mechanisms ruling bounded rational cognition and decision-making.
In this dissertation the effects of the self-referentiality of economic theories and
their absorption have been experimentally analyzed through the observation of how
individuals deal with meta-theoretical information in different experimental con-
texts. The experiments which have been conduced and presented in the frame-
work of this dissertation focussed on the absorbability of economic theories of full
rationality.
The effects of providing bounded rational individuals with meta-theoretical in-
formation of full rationality have been in particular observed in the experimental
settings of p-guessing games and of informational cascades. Both settings have sev-
eral attractive characteristics for testing theory absorption among bounded rational
decision-makers. They namely permit the effects of rationality to disentangle from
social preferences and have at the same time a very simple economic interpretation.
Further, both the guessing game and herding behaviour represent interactive settings
in which the individuals, in order to achieve a satisficing result, have to anticipate or
interpret the others’ behaviour and in which individually optimal behaving does not
per se ensure success.
In general, the two experiments provide evidence that, although a significant im-
pact of revealing theoretical information to the experimental subjects was observed,
the hypothesis of full and complete theory absorption cannot be acknowledged as
setting transcendent. Instead, the experimental findings corroborate the idea that in-
dividuals reflect the theoretical statements on themselves and on the situation they
face and adapt them to their own bounded rationality as well as to the supposed
bounded rationality and boundedly rational absorption capabilities of the others. In
interactive settings, similar to the ones experimentally tested, strategic uncertainty
about the behaviour of the others could be revealed to play a central role in orienting
individual behaviour and to definitely inspire the modalities of theory absorption by
bounded rational decision-makers.
Conclusion
The phenomenon of referring is pervasive and regards all fields of human thought
and activity, so much so that it appears to be an inescapable basis of all that can
be thought, conceptualized and expressed. Reflexivity can affect the dynamics of
social and economic systems at different levels because it concerns the social reality
per se, but it also represents an essential feature of the social sciences and social
research.
The focussing on the recursivity and self-referentiality of economic theories in-
tends to investigate the effects the knowledge and acceptance of a theory among the
acting individuals in an economic system who provide the dynamics and develop-
ment of that system.
For this reason the present analysis has deepened the modalities by means of
which real economic actors perceive the recursive character of economic theoriz-
ing, the conditions under which economic theories affect in a self-referential way
the economic actors’ behaviour and the possibility of empirically testing the self-
referentiality of economic theories.
In doing that it has relied on the notion of “theory absorption,” whose implica-
tions have been discussed considering the boundaries that are posited to the sub-
jective rationality of the economic actors. A theory is said to be fully absorbable
whenever its own acceptance by all of the individuals belonging to a certain popu-
lation does not question its predictive validity. This accounts for strategic equilibria
and can be related to the logic underlying convergence of behaviour and the stability
of equilibrium outcomes.
While assuming the economic actors to be perfectly rational, a theory of ratio-
nal choice will be completely and universally absorbed. Allowing for the bounded
rationality of the individuals requires that a certain margin be left for the individual
who is adapting to the theory’s prescriptive.
Among the requisites for a theory to be absorbed are, in particular, its under-
standing by the individuals and its compatibility with their subjective beliefs and
mental representations. A theory is absorbed by an individual if that individual in-
ternalizes it in her own mental models and chooses to act according to its logical
content. In interactive contexts, theory absorption will also be strongly related to the
supposed mental models of the others. These have been considered to be elaborated
183
184 Conclusion
by means of introspection. Depending on the number of individuals – from one to

all – who follow its prescriptions and are satisfied with the result, it can be distin-
guished among unilaterally-, partially-, and fully-absorbable theories, in particular.
Striving toward a better understanding of the mechanisms on which theory ab-
sorption relies has been interpreted as one of the possible approaches for defining
bounded rational expectations, bounded rational best replies and formulating more
realistic economic forecasts.
The effects of the self-referentiality of economic theories and their absorption
have been experimentally analyzed by observing how individuals deal with meta-
theoretical information, i.e. theoretical information about the experimental situation
faced by the respondents. For this purpose, two experiments have been conducted
focussing respectively on the absorbability of the guessing-game theory and the
herding behaviour theory.
The experiments provided evidence that, although a significant impact of reveal-
ing theoretical information to the experimental subjects was observed, the hypoth-
esis of full theory absorption (in the sense of perfect compliance to the theory’s
prescriptive) could not be acknowledged. Instead, the experimental findings corrob-
orate the idea that individuals reflect the theoretical statements on themselves and
on the situation they face and adapt them to their own bounded rationality as well as
to the supposed bounded rationality and boundedly rational absorption capabilities
of the others. In interactive settings, like the ones experimentally tested, strategic
uncertainty about the behaviour of the others could be revealed to play a central role
in orienting individual behaviour and to decisively inspire the modalities of theory
absorption by bounded rational decision-makers.
As potentially any economic theory can yield recursive effects and be absorbed
for the resolution of a concrete problem, the analysis of the self-referentiality of
economic theories and their absorption can be approached either focussing on the
reflexive effects of theories, which rely on the assumption of fully rational economic
actors, or on those of theories encompassing individuals’ bounded rationality. This
study conceived the inquiry of the absorbability of theories of full rationality as a
natural first step, and favoured it above beginning from observing the absorption of
theories of bounded rationality. Testing the absorbability of theoretical principles
of bounded rational behaviour could constitute an interesting development of the
analysis.
Besides tackling the pure speculative intent in exploring the psychology of choice
by testing if theories are accepted by bounded rational decision-makers and if their
predictive content survives such test, this research program could yield several use-
ful applications.
It could for example enhance the development of effective decision support
systems, teaching methods and training procedures which enjoy a broad accep-
tance among real (bounded rational) decision-makers. A better understanding of the
modalities by which a theory can be applied for the resolution of a practical problem
could permit the development of effective debiasing procedures which would enable
some mental blockades that are responsible for biased judgements and decisions to
be overcome.
Conclusion 185
Furthermore, the findings on the absorbability of theoretical prescriptions could

be applied to policy advising. On their basis of these findings, advice, that is fully
workable (thus absorbable) by at most bounded rational actors, could be inferred.
Hereby, absorbable advice should insist on the mechanisms which rule human cog-
nition and guide decision-making and should acknowledge both the boundedness of
its addressee and that of the behaviours on which it ought to be ruled.
Appendix

of Guessing Game Theory
Welcome to this experiment! Thank you for your participation.

You are member in a group, consisting of 8 persons. All of your group members
have these same instructions you have.
In each round, each person of your group has to choose a number between 0 and
100 (0 and 100 are possible as well). You can choose each number you like. Please
note, that numbers with more than 2 decimals are excluded. The chosen numbers of
your group members will remain unknown to you.
You can earn points in each round. The points you earn depend on how close
your number is to a target number. The closer your number to a target number, the
higher your payoff in points.
You can calculate you payoff in points this way:
target number
Earned points per round = 50–2, 5 · | your chosen number-q · (group average + d)
| distance
At the beginning of each round all group members choose a number simultane-
ously. The target number (modified group average) is the average of the numbers
your groups’ choice, added by a constant d, all multiplied by q.
The values you need can be calculated as follows:
1.1 Example
x1 = your chosen number

x2 = the number chosen by the 2nd group member
x3 = the number chosen by the 3rd group member
187
188 Appendix
...
x8 = the number chosen by the 8th group member
The average is determined by:

x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8
8
The target number is determined by: q∗ (average + d)
while q and d vary in every round and will be shown to you on the PC screen.
The target number might be bigger or smaller than your chosen number All what
matters for the calculation of your payoff is the distance between your chosen num-
ber and the target number, that’s why we take the absolute value of this difference.
Therefore the distance is always positive and determined by:
|your chosen number-target number|
Your payoff per round is: 50 points – 2, 5 · distance
If the distance time is 2.5 bigger than 50 (it means, your earnings should be
negative) you will still receive 0 points. You might not earn additional points but
you will never loose points.
At the end of each round you will receive information about your chosen number,
your earnings and the average of all chosen numbers of your group. Since q and d
vary in every round, please after each round choose your number again/please after
each round think again about the number you want to choose. After you played 9
rounds, you’ll receive recapitulating information about your payoffs. The transac-
tion course is 1 point = 1 Cent. Payoffs are accumulated over all rounds and paid in
cash and privately at the end of the experiment.
[Instructions for treatment 3]
Remember, all of your group members have the same information like you. All
know as well that all participants received the same information.
[In treatment 1 and 2 respective all and half of the participants received following
“additional tips”].
1.2 Additional Tips
1. To earn as many points as possible, you have to guess which number results
from the calculation “average of the numbers your groups’ choice added by d, all
multiplied by q”. Choose then this number.
2. Any number, chosen by all of the group members won’t exceed 100. This means,
the average of all chosen numbers won’t exceed 100. The average added by d
(here, e.g. d = 50) equals 150. Multiply then this number (group average + d)
times q (here, e.g. q = 1/3) and you get 50. Therefore the number you should
choose, should not exceed 50.
1 Instructions to the Experiment on the Absorbability of Guessing Game Theory 189
3. If all members of your group will realize this, everybody else will choose a num-
ber which does not exceed 50 and therefore the average will not exceed 50. The
average plus d = 50 (that is 100) multiplied by q = 1/3 is 33.33. Therefore the
number you should choose, should not exceed 33.33.
4. If all members of your group will realize this, everybody else will choose a num-
ber which does not exceed 33.33 and therefore the average will not exceed 33.33.
The average plus d = 50 (that is 83.33) multiplied by q = 1/3 is 27.77. Therefore
the number you should choose, should not exceed 27.77 and so on.
5. If you and every group member will think this way, everybody will realize that
the number that will be chosen will not exceed 25, so that the best everybody can
do is to choose 25. All would then choose 25 ensuring themselves the maximum
payoff of 0.5e per round.
[Instructions for treatment 1]

All of your group members received as you those additional tips. They know as
well that all group members received those tips.
[Instruction for half of participants in treatment 2]
Four people (you included) in your group received those additional tips. The
other four people are just aware of the basic instructions, but not of those additional
tips. All people in your group know as well, that there are four members who re-
ceived some additional tips and four who did not.
[Instruction for the remaining half of participants in treatment 2]
All of your group members received as you those instructions. In addition, four
of them received some advices (tips) for playing this game. All people in your group
know as well, that there are four members who received some additional tips and
four who did not.
Please answer some control questions on your PC screen before starting the ex-
periment. This ensures the understanding of the rules of this experiment. Please re-
main seated during all the time. If you have questions, please raise your hand. Your
question will be answered privately. Please, remain seated after the experiment and
wait for further instructions. Thanks!
1. Assume, you chose 24, the other group members chose 17, 66, 100, 91, 73, 82,
and respectively 13.
a. Calculate the average.
b. What is the target number, if d = 0 and q = 2/3?
c. How much is the distance to your chosen number?
2. Assume, you chose 72, and the average of numbers the other group members
chose is 78.
a. Calculate the average, inclusive your chosen number.
b. What is the target number, if d = 0 and q = 2/3?
c. How much would you earn if d = 25 and q = 1/3?
190 Appendix
3. Assume, in the previous round the parameters were d = 0 and q = 1/3. Now, in
this round d = 50 and q = 1/3. Which of the following sentences is then correct?
a. Group average added by d becomes higher.
b. The target number gets smaller.
c. Nothing can be said.

of Informational Cascades’ Theory
Welcome to this experiment! Thank you for your participation.

This is an experiment in the economics of decision-making.
Although you will have to use a computer, the experiment is extremely simple.
If you follow the instructions and make careful decisions, you may earn a consid-
erable amount of money which will be paid to you cash at the end of the experiment.
Your earnings will depend partly on your decisions and partly on chance.
Think that you have to act like an entrepreneur who has to decide either to pro-
duce a new product or not.
Two situations will be equally possible: either you will sell all of the products
you produce or none of them.
In each of the 20 periods you are going to play, the computer will determine the
result of your sales (if you will sell everything or nothing). The sales result will be
the same for all of people playing this game and will be determined period after
period.
Each time you will make the right choice, you will earn 0.5e, while if you
take the wrong decision, you will earn nothing. This is summarized in the follow-
ing table:
Choice: to invest Choice: not to invest

Sales go well 0.5e 0e
Sales go bad 0e 0.5e
You will play with other subjects and you will be asked to make your choice in a
certain order which will be randomly determined and will be different in each of the
20 periods.
Further, in order to make you choice, you will receive two different kinds of
information:
The first one is private: think for example that you commissioned a survey on the
success of your new product. The agency which did the survey ensures their results
to be correct with a probability of 75%.
In the experiment, the computer will give you this information, in form of a signal
which can be equal 1 or 0.
As explained in the following table, a signal equals 1 tells you that sales will go
well in 75% of the cases while they will go bad in 25% of the cases.
2 Instructions to the Experiment on the Absorbability of Informational Cascades’ Theory 191
A signal equals 0 tells you, on the contrary, that sales will go well in the 25% of
the cases and that they will go bad in the 75%.
Signal = 1 Signal = 0
Sales go well 75% 25%
Sales go bad 25% 75%
The second information is available to all the participants (it is “public”) and con-
sists in knowing the choices of the subjects who played before you.
Therefore, the number of information you will receive depends on the order with
which you will be asked to play. Assume, for example, that you are the fifth subject
to play: in this case you know your signal and the choices of the four preceding
players. If you are the sixth player, you will know your signal and the choices of the
five preceding players and so on.
After receiving this information, you will have 10 s to decide, either to invest
or not (the time that remains you for typing your choice will be displayed in the
upper-right corner of your pc-screen).
After all of the players will have made their choice, you will be informed about
the right decision and about your profit.
You will play in this way 20 times, added by two initial trials. After having played
all periods, the profit you gained in the game (which excludes the profit you earn in
the two initial trials) will be paid you cash. You will be then allowed to leave the
laboratory.
Please, do not talk to anyone during the experiment. We ask everyone to remain
silent until the end of the experiment. If you have questions about the game, you can
ask the experimenters during the two trials by raising your hand.
Your participation in the experiment and any information about your earnings
will be kept strictly confidential.
Good luck!
[In the absorption treatment all of the participants received following “addi-
tional tips”].
2.1 Additional Tips
Think of the game carefully:

If you are the first player to move and the signal you receive is “1”, you know
that in 75% of the cases sales will go well, that is, the probability that sales will go
well, given the signal 1, is 0.75. Therefore, if you choose to invest, your chances to
win stays 75%. Since this is more than 50% (i.e. more than 0,50 probability), if the
signal you receive is “1”, you better choose to invest.
However, if you are the first player and the signal you receive is “0”, you know
that in 25% of the cases sales will go well, that is, that is, the probability that sales
will go well, given the signal “0”, is 0.25. Therefore, if you choose to invest, your
192 Appendix
chances to win stay 25%. Since this is less than 50% (i.e. less than 0,50 probability),
if the signal you receive is “0”, you better choose not to invest.
If you are not the first player, the first player will think the same and therefore
you can expect him/her:
- to invest, if his/her signal is “1”;
- not to invest, if his/her signal is “0”.
The 2nd, the 3rd or the 4th player will think the same, so that you can expect each
of them:
- to invest, if his/her signal is “1”;
- not to invest, if his/her signal is “0”.
Assume you are the 5th player, in addition to the signal you receive, you can observe
what the players before you have chosen.
In particular, you can be faced with one of the following scenarios:
- all of the 4 players before you have chosen to invest;
- 3 of them have chosen to invest, only one of them not to invest;
- 2 of them have chosen to invest, the other 2 not to invest;
- only one of them has chosen to invest, the other 3 not to invest;
- they have all chosen not to invest.
E.g. if all the 4 players before you have chosen to invest, they all probably received
a signal equal “1”.
Therefore, if you receive a signal equal “1”, you better choose to invest, too.
BUT, what would you do, if you receive a signal equal “0”?
You can think that, if the signal was “1” in 4 of the 5 cases, and “0” just in your
one case,
the probability that given all signals sales will go well can be calculated
this way:
1 1 1 1 1
· 0.75 + · 0.75 + · 0.75 + · 0.75 + · 0.25
5 5 5 5 5
This is because, in the first fifth of cases (for the 1st out of 5 players) the probability
that given the signal sales will go well is 0.75, so that it counts for 15 · 0.75;
For the 2nd fifth of cases(for the 2nd out of 5 players) the probability that given
the signal sales will go well is further 0.75. That adds to the previous probability
and makes in total: 15 · 0.75 + 15 · 0.75;
The same adds for the 3rd player (that is 15 · 0.75 + 15 · 0.75 + 15 · 0.75) and the 4th
player (that is 15 · 0.75 + 15 · 0.75 + 15 · 0.75 + 15 · 0.75)
In the fifth case (your case) your signal, which equals “0”, tells you that the
probability for the sales to go well is 0,25, so that 15 · 0.25 adds.
Overall, the probability that given all signals sales will go well is:
1
5 · 0.75 + 15 · 0.75 + 15 · 0.75 + 15 · 0.75 + 15 · 0.25 = 0.65 > 0.50
2 Instructions to the Experiment on the Absorbability of Informational Cascades’ Theory 193
Since this probability is bigger than 0.50, your chances to win if you invest, despite
of your own signal, stays higher than 50%.
Therefore, you can better choose to invest, even if your own signal is “0”.
Imagine now that you are the 6th player and that 3 players before you have chosen
to invest and 2 not to invest.
In this case, if your signal is 1, you better choose to invest, since the probability
that given all signals sales will go well is bigger than 0.50,
Whilst if your signal is 0, the probability that given all signals sales will go well
is exactly 0.50, since 15 ·0.75+ 15 ·0.75+ 15 ·0.75+ 15 ·0.25+ 15 ·0.25+ 15 ·0.25 = 0.50
This means, both if you choose to invest or not, you have the same chances to
win. You can therefore choose at random, as tossing a coin.
From all that you can think of a simple rule for playing this game:
1. if an equal number of players before you have chosen to invest or not to invest
(e.g. 3 invested and 3 not), than follow your signal;
2. if among your predecessors there is one more player that have chosen to invest
(this is e.g. the case when 4 players invested and 3 not) and your signal is 1, then
choose to invest. Respective, if one more player have chosen not to invest, and
your signal is 0, then choose not to invest;
3. invest, regardless of your signal, if among your predecessors two more players
before you have chosen to invest (this is e.g. the case when 5 players invested and
3 not). Respective, if two more players have chosen not to invest, do not invest,
regardless of your signal.
Literature
Adolphs, R. (2003). Cognitive neuroscience of human social behaviour. Nature Review, 4, 165–
178.
Agnoli, F., & Krantz, D. H. (1989). Suppressing natural heuristics by formal instruction: The case
of the conjunction fallacy. Cognitive Psychology, 15(4), 439–449.
Albert, H. (Ed.). (1972). Theorie und realität, Tübingen: J. C. B. Mohr (Paul Siebeck).
Allais, M. (1953). Les comportement de l’homme rationell devant le risque: Critique des postulats
at axiomes de l’ecole americane. Econometrica, 21, 503–46.
Allsopp, L., & Hey, J. D. (2000). Two experiments to test a model of herd behaviour. Experimental
Economics, 3, 121–136.
Anderson, L. R. (2001). Payoff effects in information cascade experiments. Economic Inquiry,
39(4), 609–615.
Anderson, L. R., & Holt, C. A. (1997). Information cascades in the laboratory. American Economic
Review, 87, 847–862.
Arrow, K. J. (1982). Risk perception in psychology and economics. Economic Inquiry, 20, 1–9.
Avery, C., & Zemsky, P. (1998). Multidimensional uncertainty and herd behavior in financial mar-
kets. The American Economic Review, 88(4), 724–748.
Bannerjee, A. V. (1992). A simple model of herd behaviour. Quarterly Journal of Economics, 107,
797–817.
Bartlett, S. J. (1987). Varieties of self-reference. In S. J. Bartlett, & P. Suber (Eds.) Self-reference.
Reflections on reflexivity (pp. 5–28). Dordrecht: Martinus Nijhoff Publishers.
Bartlett, S. J. (1992). Reflexivity. A source-book in self-reference. North-Holland: Elsevier Science
Publishers B.V.
Bartlett, S. J., & Suber, P. (Eds.). (1987). Self-reference. Reflections on reflexivity. Dordrecht: Mar-
tinus Nijhoff Publishers.
Barwise, J., & Etchemendy, J. (1987). The Liar – An essay on truth and circularity. New York:
Oxford University Press.
Bateson, G. (1972). The logical categories of learning and communication. In G. Bateson (1972).
Steps to an ecology of mind. New York: Ballantine.
Bateson, G. (1991). A social scientist views the emotions. In R. Donaldson (Ed.) Gregory Bateson:
Further steps to an ecology of mind. New York: Harper & Row.
Beard, R. T., & Beil, R. (1994). Do People Rely on the Self-interested Maximization of Others?:
An Experimental Test. Management Science, Vol. 40, No. 2 (Feb., 1994), 252–262.
Becker, G. S. (1991). A note on restaurant pricing and other examples of social influences on price.
Journal of Political Economy, 99(5), 1109–1116.
Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity and social history. Games and
Economic Behavior, 10, 122–142.
195
196 Literature
Berger, P. L., & Luckmann, T. (1966). The social construction of reality: A treatise in the sociology
of knowledge. Garden City, NY: Anchor Books.
Berninghaus, S., Güth, W., & Kliemt, H. (2003a). From teleology to evolution: Bridging the gap
between rationality and adaptation in social explanation. Journal of Evolutionary Economics,
13(4), 385–410.
Berninghaus, S., Güth, W., & Kliemt, H. (2003b). Reflections on equilibrium: Ideal rationality and
analytic decomposition of games. Homo Oeconomicus, 20, 257–302.
Bernoulli, D. (1738). Exposition of a new theory on the measurement of risk. Econometrica,
22, 23–36.
Bikhchandani, S., Hirshleifer, D., & Welch, I. (1992). A theory of fads, fashion, custom, and cul-
tural change as informational cascades. Journal of Political Economy, 100, 992–1026.
Bikhchandani, S., & Sharma, S. (2001). Herd behavior in financial markets. IMF Staff Papers,
47(3), 279–310.
Binmore, K. (1987). Modeling Rational Players, Part I. Economics and Philosophy, October 1987,
3, pp. 179–214.
Bolander, T. (2002). Self-reference and logic. Phi News, 1, 9–44.
Bolander, T. (2003). Logical theories for agent introspection. PhD thesis, Informatics and Mathe-
matical Modelling (IMM), Technical University of Denmark.
Bolle, F. (1979). Das problem des optimalen Stoppens: modelle und experimente. Frankfurt:
Peter Lang.
Bolton, G. E. (1998). Bargaining and dilemma games: From laboratory data towards theoretical
synthesis. Experimental Economics, 1, 257–281.
Bombach, G. (1962). Über die möglichkeit wirtschaftlicher voraussagen. Kyklos, XV, 29–67.
Bosch-Domenech, A., Montalvo, J. G., Nagel, R., & Satorra, A. (2002). One, two, (three),
infinity. . .: Newspaper and lab beauty-contests experiments. The American Economic Review,
92(5), 1687–1701.
Bosse, L. (1957). Über die möglichkeiten und den Nutzen von kurzfristigen Wirtschaftsprognosen.
Weltwirtschaftliches Archiv, 79, 65–83.
Buck, R. C. (1963). Reflexive predictions. Philosophy of Science, 30(4), 359–369.
Bush, R., & Mosteller, F. (1955). Stochastic models of learning. New York: Wiley.
Calandro, J., Jr. (2004). Reflexivity, business cycles, and the new economy. The Quarterly Journal
of Austrian Economics, 7(3), 45–69.
Camerer, C. (1995). Individual decision making. In J. H. Kagel, & Roth, A. (Eds.), The handbook
of experimental economics (pp. 587–616). Princeton: Princeton University Press.
Camerer, C. F. (2003a). Behavioral game theory. New York: Russel Sage Foundation; Princeton,
NJ: Princeton University Press.
Camerer, C. F. (2003b). Behavioural studies of strategic thinking in games. TRENDS in Cognitive
Sciences, 7(5), 225–231.
Camerer, C. F., Ho, T.-H., & Chong, J.-K. (2001). Sophisticated experience-weighted attraction
learning and strategic thinking in repeated games. Journal of Economic Theory, 104, 137–188.
Campbell, D. T. (1969). Reforms as experiments. American Psychologist, 24, 409–429.
Carey, S. (1985). Conceptual differences between children and adults. Mind & Language, 3(3),
167–181.
Carnap, R. (1942). Introduction to semantics, Cambridge, MA: Harvard University Press.
Carnap, R. (1956). Meaning and necessity. Chicago: University of Chicago Press.
Cass, D., & Shell, K. (1983). Do sunspots matter? Journal of Political Economy, 91(2), 193–227.
Cassel, S. (2000). Politikberatung und Politikerberatung – Eine institutionenökonomische Analyse
der wissenschaftlichen Beratung der Wirtschaftspolitik. Bern: Verlag Paul Haupt.
Çelen, B., & Kariv, S. (2004). Distinguish information cascades from herd behavior in laboratory.
American Economic Review, 94(3), 484–498.
Chase, V. M., Hertwig, R., & Gigerenzer, G. (1998). Visions of rationality. Trends in Cognitive
Sciences, 2(6), 206–214.
Chew, S. H., & McCrimmon, K. R. (1979). Alpha-nu choice theory: An axiomatization of expected
utility. Working paper, 669. University of British Columbia Faculty of Commerce.
Literature 197
Church, A. (1951). The need for abstract entities in semantic analysis. Proceedings of the American
Academy of Arts and Sciences, 80, 100–112.
Cipriani, M., & Guarino, A. (2005). Herd behavior in a laboratory financial market. American
Economic Review, American Economic Association, 95(5), 427–1443.
Coase, R. (1974). The lighthouse in economics. Journal of Law and Economics, 17(2), 357–376.
Cole, W. G. (1989). Understanding Bayesian reasoning via graphical displays. Paper presented in
CHI’89 Proceedings, Austin, TX: ACM Press.
Copeland, T., Koller, T., & Murin, J. (2000). Valuation, New York: Wiley.
Cournot, A. (1838). Recherches sur les principes mathématiques de la théorie des richesses, Paris.
Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Englewood Cliffs, N.J.:
Prentice-Hall.
Cyert, R. M., Simon, H. A., & Trow, D. B. (1956). Observation of a business decision. Journal of
Business, 29, 237–248.
Dacey, R. (1976). Theory absorption and the testability of economic theory. Zeitschrift für Na-
tionalökonomie, 36, 247–267.
Dacey, R. (1981). Some implications of ‘theory absorption’ for economic theory and the economics
of information. In J. C. Pitt (Ed.), Philosophy in economics, Dordrecht: D. Reidel Publishing
Company.
Davidson, D. (2001). Inquiries into truth and interpretation (2nd ed.) New York: Oxford Univer-
sity Press.
Davidson, D., & Marschak, J. (1959). Experimental test of a stochastic decision theory. In Mea-
surement: definitions and theories. Churchman, C. W., & Ratoosh, P. (eds). New York: Wiley.
233–69
Davidson, D., Suppes, P., & Siegel, S. (1957). Decision making: an experimental approach. Stan-
ford, Calif.: Stanford University Press.
Davis, J. B., & Klaes, M. (2003). Reflexivity: Curse or cure? Journal of Economic Methodology,
10(3), 329–352.
Dawes, R. (1988). Rational choice in an uncertain world. San Diego: Harcourt, Brace, Jovanovich.
Demandt, A. (2001). Ungeschehene Geschichte. Ein Traktat über die Frage: Was wäre geschehen,
wenn. . .?, 3. erw. Aufl., Vandenhoeck und Ruprecht, Kleine Reihe V & R 4022, Göttingen.
Dennett, D. (1987). The intentional stance (Reprint ed.). Cambridge, MA: MIT Press.
Douglas, G. (1996). What have we learn from social learning? European Economic Review, 40,
617–628
Duffy, J., & Nagel, R. (1997). On the robustness of behaviour in experimental ‘beauty contest’
games. The Economic Journal, 107(445), 1684–1700.
Duncker, K. (1935). On problem solving (L. S. Lees, Trans.). Psychological Monographs, 58 (orig-
inal work published 1935).
Edwards, W. (1953a). Experiments on economic decision-making in gambling situations. Abstract.
Econometrica, 21:349–50.
Edwards, W. (1953b). Probability preferences in gambling. American Journal of Psychology,
66:349–64.
Egidi, M. (1992). Organizational learning, problem solving and the division of labour. In H. Si-
mon, M. Egidi, R. Marris, R. Viale (1992). Economics, bounded rationality and the cognitive
revolution (pp. 148–173), UK: Edward Elgar Publishing Company.
Egidi, M. (2003). Decomposition patterns in problem solving. Social Science Research Network
Electronic Library. Accepted Paper Series.
Egidi, M. (2005). From bounded rationality to behavioral economics. EconWPA, Experimental
0507002.
Egidi, M., & Narduzzo, A. (1997). The emergence of path-dependent behaviors in cooperative
contexts. International Journal of Industrial Organization, 5, 677–709.
Einhorn, H. J. (1982). Learning for experimence and suboptimal rules in decision-making. In
D. Kahnemann, P. Slovic, A. Tversky (Eds.), Judgement under uncertainty (268–283). Cam-
bridge: Cambridge University Press.
198 Literature
Einstein, A. (1905, March). On a heuristic point of view concerning the generation and transfor-
mation of light (Concerning the generation and transformation of light from a heuristic point
of view). Annalen der Physik, 17, 132–148.
Elveton, R. O. (Ed.) (2000). The phenomenology of husserl: Selected critical readings. Seattle:
Noesis Press.
Fellner, G., Güth, W., & Martin, E. (2006a). Satisficing or optimizing? – An experimental study.
Papers on Strategic Interaction, No. 11–2006, Max Planck Institut für Ökonomik, Jena.
Fellner, G., Güth, W., & Martin, E. (2006b). Task transcending satisficing – An experimental study.
Papers on Strategic Interaction, No. 09–2006, Max Planck Institut für Ökonomik, Jena.
Ferguson, N. (Ed.). (1999). Virtuelle Geschichte. Historische Alternative im 20. Jahrhundert (Ger-
man Trans.). Darmstadt: Primus.
Festinger, L. (1957). A theory of cognitive dissonance. Stanfort, CA: Stanford University Press.
Fiore, A., & Morone, A. (2007). A simple note on informational cascades. Economics Discussion
Papers, 2007–21, Kiel Institute for the World Economy.
Fiore, A., Morone, A., & Sandri, S. (2007). On the absorbability of herd behaviour and informa-
tional cascades: An experimental analysis. Dresden Discussion Paper in Economics, No. 13/07.
Fischbacher, U. (1998). Z-tree: A toolbox for readymade economic experiments. University of
Zurich.
Fischer, H. R. (Ed.). (1995). Die Wirklichkeit des Konstruktivisms – Zur Auseinandersetzung um
ein neues Paradigma (pp. 47–71). Heidelberg: Carl-Auer-Systeme-Verlag.
Fischoff, B. (1982a). For those condemned to study the past: Heuristics and biases in hindsight.
In D. Kahnemann, P. Slovic, A. Tversky (Eds.), Judgement under uncertainty (pp. 335–352).
Cambridge: Cambridge University Press.
Fischoff, B. (1982b). Debiasing. In D. Kahnemann, P. Slovic, & A. Tversky (Eds.), Judgement
under uncertainty (pp. 422–442). Cambridge: Cambridge University Press.
Franz, W. (2000). Wirtschaftspolitische Beratung. Perspektiven der Wirtschaftspolitik, Band 1, Heft
1, 53–71.
Frederick, S. (2003). Time preference and personal identity. In G. Loewenstein, D. Read, &
R. Baumeister (Eds.). (2003). Time and decision: Psychological perspectives in intertempo-
ral choice. New York. Russel Sage.
Frey, B. S. (2000). Was bewirkt die Volkswirtschaftslehre? Perspektiven der Wirtschaftspolitik,
Band 1, Heft 1, 5–33.
Frey, D. (1986). Recent research on selective exposure to information. In L. Berkowitz (Ed.), Ad-
vances in experimental social psychology (Vol. 19, pp. 41–80). NY: Academic Press.
Friedman, D., & Sunder, S. (1994). Experimental methods, a primer for economists. Cambridge:
Cambridge University Press.
Friedman, M. (1953). Essays in positive economics. Chicago: Chicago University Press.
Friedman, M., & Savage, L. J. (1952). The expected utility hypothesis and the measurability of
utility. Journal of Political Economy, 60, 6.
Fulda, E., Lehmann-Waffenschmidt, M., & Schwerin, J. (1998). Zwische Zufall und
Notwendigkeit – Zur Kontingenz ökonomischer Prozesse aus theoretischer und historischer
Sicht. In G. Wegner, & J. Wieland (Ed.), Formelle und informelle Institutionen. Genese, Inter-
aktion und Wandel, Marburg: Metropolis.
Geanakoplos, J. (1989). Game theory, partitions, and applications to speculation and consensus.
Cowles Foundation Discussion Paper, 914, Yale University.
Gebotys, R. J., & Claxton-Oldfield, S. P. (1989). Errors in the quantification of uncertainty: a
product of heuristics or minimal probability knowledge base? Applied Cognitive Psychology,
3, 237–250.
Gigerenzer, G., & Goldstein, D. G. (1996). Reasoning the fast and frugal way: Models of bounded
rationality. Psychological Review, 103, 650–669.
Gigerenzer, G., & Goldstein, D. G. (2002).Models of ecological rationality: The recognition heuris-
tic. Psychological Review, 109(1), 75–90.
Gigerenzer, G., Hell, W., & Blank, H. (1988). Presentation and content: The use of base rates as
a continuous variable. Journal of Experimental Psychology: Human Perception and Perfor-
mance, 14, 513–525.
Literature 199
Gigerenzer, G., & Selten, R. (Eds.). (2001). Bounded rationality: The adaptive toolbox.
Cambridge, MA: MIT Press.
Gigerenzer, G., & Todd, P. M. (1999). Fast and frugal heuristics: The adaptive toolbox. In G.
Gigerenzer, P. Todd, & the ABC Research Group (Eds.). (1999). Simple heuristics that make
us smart (pp. 3–34). Oxford: Oxford University Press.
Gigerenzer, G., Todd, P., & the ABC Research Group (Eds.). (1999). Simple heuristics that make
us smart. Oxford: Oxford University Press.
Gilbert, D. T. (2002). Inferential Correction. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.),
Heuristics and biases (pp. 167–184). New York: Cambridge University Press.
Greiner, B. (2004). The online recruitment system orsee 2.0 – A guide for the organization of
experiments in economics. Working Paper Series in Economics 10, University of Cologne,
Department of Economics.
Grether, D. M., & Plott, C. R. (1979). Economic theory of choice and the preference reversal
phenomenon. American Economic Review, 69, 623–638.
Groener, R. (Ed.). (1983). Methods of heuristics. Hillsdale, NJ: Erlsbaum.
Grünbaum, A. (1956). Historical determinism, social activism and predictions in the social sci-
ences. British Journal for the Philosophy of Science, 7(27), 236–240.
Grunberg, E., & Modigliani, F. (1954). The predictability of social events. Journal of olitical Econ-
omy, 62, 465–478.
Gul, F. (1991). A theory of disappointment in decision-making under uncertainty. Econometrica,
59(3), 667–686.
Güth, W. (2000). Boundedly rational decision emergence – A general perspective and some selec-
tive illustrations. Journal of Economic Psychology, 21, 433–458.
Güth, W. (2006). Satisficing in portfolio selection – Theoretical aspects and experimental tests.
Discussion Papers on Strategic Interaction, Max Planck Institute for Economics, Jena.
Güth, W., & Kliemt, H. (2000). From full to bounded rationality – The limits of unlimited rational-
ity. ZiF – Mitteilungen Special 2000, Zentrum für interdisziplinäre Forschung der Universität
Bielefeld, 1–15.
Güth, W., & Kliemt, H. (2004a). Bounded rationality and theory absorption. Homo Oeconomicus,
21(3/4), 251–540.
Güth, W., & Kliemt, H. (2004b). Perfect or bounded rationality?: Some facts, speculations and
proposals. Analyse und Kritik: Zeitschrift für Sozialtheorie, 26(2), 364–381.
Güth, W., Kocher, M., & Sutter, M. (2002). Experimental’beauty contests’ with homogeneous and
heterogeneous players and with interior and boundary equilibria. Economic Letters, 74, 219–
228.
Güth, W., Levati, M. V., & Ploner, M. (2006). Is satisficing absorbable? – An experimental study.
Papers on Strategic Interaction, No. 09–2006, Max Planck Institut für Ökonomik, Jena, 2006,
forthcoming in Journal of Behavioral Finance.
Güth, W., Rolf Schmittberger & Bernd Schwarze (1982). ‘An Experimental Analysis of Ultimatum
Bargaining’, Journal of Economic Behavior and Organization, 3, 367–388
Haberlandt, K. (1998). Human memory: Exploration and application, Boston: Allyn & Bacon.
Handlbauer, G. (1997). Self-reference in individual and social decision-making. Journal of Insti-
tutional and Theoretical Economics (JITE), 153, 762.
Hands, D. (2001). Reflections without rules: Economic methodology and contemporary science
theory. Cambridge: Cambridge University Press.
Hayek, F. (1970). Can we still avoid inflation? In R. Ebeling (Ed.), The austrian theory of the trade
cycle and other essays. Auburn: Ludwig von Mises Institute.
Hazlitt, H. (1996). Economics in one lesson. San Francisco: Laissez Faire Books
Hedgeworth (1981). Mathematical Psychics (reprint of 1961), London.
Heisenberg, W. (1927). Über den anschaulichen Inhalt der quantentheoretischen Kinematik und
Mechanik. Zeitschrift für Physik, 43, 172–198.
Heisenberg, W. (2000). Physik und philosophie (6th ed.), (1st ed., 1959), Stuttgart: Hirzel Verlag.
Hempel, C. G. (1960). Inductive inconsistencies. Synthese, 12, 439–469.
200 Literature
Hempel, C. G. (1962). Deductive-nomological vs. statistical explanation. In H. Feigl, &

G. Maxwell (Eds.), Minnesota studies in the philosophy of science (Vol. III, pp. 98–169). Min-
neapolis: University of Minnesota Press.
Hey, J. D. (1991). Experiments in economics. Oxford: Blackwell.
Hey, J. D. (1994). Experimental economics. Heidelberg: Physica Verlag.
Higgins, E. T. (1996). Knowledge activation: accessibility, applicability, and salience. In
E. T. Higgins, A. Kruglanski (Eds.), Social psychology: Handbook of basic principles
(pp. 133–168). New York: Guilford Press.
Hintikka, J., & Pietarinen, J. (1966). Semantic information and inductive logic. In J. Hintikka, &
P. Suppes (Eds.), Aspects of inductive logic (pp. 96–112), Amsterdam: North Holland.
Hintikka, J., & Suppes, P. (Eds.). (1966). Aspects of inductive logic. Amsterdam: North Holland.
Ho, T. H., Camerer, C., & Weigelt, K. (1998). Iterated dominance and iterated best response in
experimental ‘p-beauty contests’. The American Economic Review, 88(4), 947–969.
Holmes, O. W. (1980). Ralph Waldo Emerson. New York: Chelsea House Pub.
Hoyenga, K., & Hoyenga, K. (1988). Psychobiology: The neuron and behaviour. Pacific Grove,
CA: Brooks/Cole Publishing.
Huck, S., & Oechssler, J. (2000). Informational cascades in the laboratory: Do they occur for the
right reasons? Journal of Economic Psychology, 21(6), 661–671.
Hung, A., & Dominitz, J. (2004). Homogeneous actions and hetergeneous beliefs: Experimental
evidence on the formation of information cascades. Econometric Society 2004 North American
Winter Meetings.
Kagel, J. H. & Roth, A. (Eds.). (1995). The handbook of experimental economics. Princeton:
Princeton University Press.
Kahneman, D. (2002, December). Maps of bounded rationality: A perspective on intuitive judge-
ment and choice. Nobel Prize Lecture.
Kahnemann, D. (2003, December). Maps of bounded rationality: Psychology for behavioral eco-
nomics. American Economic Review, 1449–1475.
Kahneman, D., & Frederick, S. (2002). Representativeness revisited: Attribute substitution in in-
tuitive judgment. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.), Heuristics & biases: The
psychology of intuitive judgment (pp. 49–81). New York: Cambridge University Press.
Kahneman, D., & Tversky, A. (1974). Judgement under uncertainty: Heuristics and biases. Science,
185, 1124–1130.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econo-
metrica, 47(19, 33), 263–291.
Kahneman, D., & Tversky, A. (1983). Extension versus intuitive reasoning: The conjunction fallacy
in probability judgment. Psychological Review, 90, 293–315.
Kahneman, D., & Tversky, A. (Eds.). (2000). Choices, values and frames. New York: Cambridge
University Press and the Russell Sage Foundation.
Kahnemann, D., Slovic, P., & Tversky, A. (1982). Judgement under uncertainty. Cambridge:
Cambridge University Press.
Keynes, J. M. (1936). The general theory of employment, investment and money. London:
Macmillan.
Kincheloe, J. (1991). Teachers as researchers: Qualitative inquiry as a path of empowerment.
London: Falmer Press.
Kliemt, H. (2001). Rationality and reality. Kyklos, 54, 309–316.
Kripke, S. (1975). Outline of a theory of truth. Journal of Philosophy, 72, 690–712.
Krippendorff, K. (1986). Web dictionary of cybernetics and systems. A Dictionary of Cybernetics.
Retrieved March 21, 2006, from http://pespmc1.vub.ac.be/Asc/Kripp.html
Lacey, A. R. (1996). A dictionary of philosophy (3rd ed.). London, UK: Routledge.
Lehmann-Waffenschmidt, M. (1990). Predictability of economic processes and the morgenstern
paradox. Schweizerische Zeitschrift für Volkswirtschaft und Statistik, 2, 147–160.
Lehmann-Waffenschmidt, M. (1996). Limitations of social forecasting. Wissenschaftliche
Zeitschrift der Technischen Universität Dresden, 45(4), 43–47.
Literature 201
Lehmann-Waffenschmidt, M. (2002). Neuer Fokus Viabilität. Zur Bedeutung des (radikalen) Kon-
struktivismus für die Wirtschaftswissenschaften und die Nachhaltigkeitsdebatte. Ökologisch
Wirtschaften, 6, 23–26.
Lehmann-Waffenschmidt, M. (2006a). Konstruktivismus und Evolutorische Ökonomik. In
G. Rusch, (Ed.), Konstruktivistische Ökonomik (pp. 27–54). Marburg: Metropolis-Verlag.
Lehmann-Waffenschmidt, M. (2006b). Self-referential optimal delays when reactions are delayed.
Dresden Discussion Paper Series in Economics. Technical University of Dresden.
Lehmann-Waffenschmidt, M., & Reina, L. (2003). Coalition formation in multilateral negotiations
with a potential for logrolling: An experimental analysis of negotiators’ cognition processes.
Dresden Discussion Papers Series in Economics, No. 17, Dresden University of Technology.
Levi, I. (1967a). Gambling with truth, New York: Alfred Knopf.
Levi, I. (1967b). Information and inference. Synthese, 17, 369–391.
Lichtenstein, S., & Slovic, P. (1971). Reversal of preferences between bids and choices in gambling
decisions. Journal of Experimental Psychology, 89, 46–55.
Loewenstein, G. (1996). Out of control: Visceral influences on behavior. Organizational Behavior
and Human Decision Processes, 65, 272–292.
Loomes, G., & Sudgen, R. (1982). Regret theory: An alternative theory of rational choice under
uncertainty. Economic Journal, 92, 805–825.
Lorenz, K. (1973). Die Rückseite des Spiegels. München: R. Piper & Co. Verlag.
Luger, F. G. (Ed.). (1995). Computation and intelligence: Collected readings. Cambridge, MA:
AAAI Press/MIT Press.
Luhmann, N. (1984). Soziale Systeme. Grundriss einer allgemeinen Theorie. Frankfurt: Surhkamp.
Mackie, J. L. (1973). Truth, probability and paradox. Oxford: Oxford University Press.
Mackinnon, L. A. K. (2003). The Social Construction of Economic Man: The Genesis, Spread,
Impact and Institutionalisation of Economic Ideas. Doctoral Thesis submitted at the University
of Queensland in March 2006 (available at http://search.arrow.edu.au/09.09.2008).
March, J. G. (1994). A primer on decision making. New York: Free Press.
Margert, A. W. (1929). Morgenstern on the methodology of economic forecasting. Journal of Po-
litical Economy, 37(3), 312–339.
Mark, T. K., & Eysenck, M. W. (2005). Cognotove psychology: A student’s handbook. UK: Psy-
chology Press.
Martin, R. L. (1967). Toward a solution to the liar paradox. The Philosophical Review 76, 279–311.
Martin, R. L. (1992). On non-translational semantics. In S. J. Bartlett (Ed.), Reflexivity. A source-
book in self-reference. North-Holland: Elsevier Science Publishers B. V.
Marx, R. (March, 2006). Die Prognostizierbarkeit des wirtschaftlichen Verhaltens – eine theo-
retische Analyse aus radikalkonstruktivistischer Perspektive und eine experimentelle Fallstudie
zu Selbstreferentialität von Prognosen und Theorien. Dissertation submitted for a diploma
at the Technische Universität Dresden under the supervision of Prof. Dr. M. Lehmann-
Waffenschmidt.
Maturana, H. R., & Varela, F. J. (1973). Autopoiesis: The organization of the living. In
H. R. Maturana, & F. G. Varela (Eds.), Autopoiesis and cognition. Dordrecht, Netherlands:
Reidel.
Maturana, H. R., & Varela, F. G. (1980). Autopoiesis and cognition. Dordrecht, Netherlands:
Reidel.
Maturana, H. R., & Varela, F. J. (1987). The tree of knowledge: The biological roots of human
understanding. Boston: Shambhala.
May, K. (1954). Intransitivity, utility, and the aggregation of preference patterns. Econometrica,
22:1–13.
McKelvey, R., & Palfrey, T. (1992). An Experimental Study of the Centipede Game. Econometrica,
July 1992, 60(4), 803–36.
McClelland, J. L., & Rumelhart, D. E. (Eds.). (1986). Parallel distributed processing. Cambridge,
MA: MIT Press.
McNeil, B. J., Pauker, S. G., Sox, H. C., & Tversky, A. (1982). On the elicitation of preferences
for alternative therapies. New England Journal of Medicine, 306, 1259–1262.
202 Literature
Meier, A., & Slembeck, T. (1994). Wirtschaftspolitik. Ein kognitiv-evolutionärer Ansatz. München
Wien: Oldenbourg Verlag.
Mendelson, E. (1997). Introduction to mathematical logic (4th ed.). London: Chapman & Hall.
Merriam Webster’s Unabridged Dictionary (2000).
Merton, R. K. (1936). The unanticipated consequences of purposive social action. American Soci-
ological Review, 1(6), 894–904.
Mistri, M. (1997). Changing preferences and cognitive processes. In R. Viale, (Ed.), Cognitive
economics, LaSCoMES Series, I/1997.
Morgenstern, O. (1928). Wirtschaftsprognose, eine Untersuchung ihrer Voraussetzungen und
Möglichkeiten. Vienna: Julius Springer Verlag.
Morgenstern, O. (1935). Vollkommene Voraussicht und wirtschaftliches Gleichgewicht. Zeitschrift
für Nationalökonomie, 6, 337–357.
Morgenstern, O. (1972). Descriptive, predictive and normative theory, Kyklos, XXV, 699–714.
Morgenstern, O., & Schwödiauer, G. (1976). Competition and collusion in bilateral markets,
Zeitschrift für Nationalökonomie, 36, 217–245.
Morone, A., & Morone, P. (2007). Guessing game and people behaviours: What can we
learn? Mimeo.
Morone, A., Sandri, S., & Uske, T. (2008). On the absorbability of the guessing game theory – A
theoretical and experimental analysis. In A. Innocenti, & P. Sbriglia (Eds.), Games, rationality
and behaviour. Essays on behavioural game theory and experiments (pp. 161–181). Hound-
mills: Palgrave McMillan.
Mosteller, F., & Nogee, P. (1951). An experimental measurement of utility. Journal of Political
Economy, 59:371–404.
Moutier, S., & Houdé, O. (2003). Judgement under uncertainty and conjunction fallacy inibition
training. Thinking and Reasoning, 9(3), 185–201.
Mueller, D. C. (Ed.). (1997). Perspectives on public choice: A handbook. Cambridge: Cambridge
University Press.
Mumma, G. H., & Wilson, S. B. (1995). Procedural debiasing of primacy/anchoring effects in
clinical-like judgements. Journal of Clinical Psychology, 51(6), 841–853.
Nagel, R. (1995). Unraveling in guessing games: An experimental study. The American Economic
Review, 85(5), 1313–1326.
Nagel, R. (1999). A survey on experimental guessing games: A study of bounded rationality and
learning. In D. V. Budescu,, I. Erev, & R. Zwick, Games and human behavior (pp. 105–42).
London: Lawrence Erlbaum.
Nash, J. (1951). Non-cooperative games. Annals of Mathematics, 54, 286–295.
Neisser, U. (1976). Cognition and reality. Principles and implications of cognitive psychology, San
Francisco: Freemen.
Nelson, R. R., & Winter, S. (1982). An evolutionary theory of economic change. Cambridge, MA:
The Belknap Press of the Harvard University Press.
Newell, A., & Simon, H. A. (1972). Human problem solving, Englewood Cliff: Prentice Hall.
Niiluoto, I., & Tuomela, R. (1973). Theoretical concepts and hypothetico-inductive inference. Dor-
drecht: Reidel.
Peirce, C. S. (1931–1958). In Hartshorne, C., Weiss, P., & Burks, A. (Eds.), Collected papers
I–VIII. Cambridge, MA: Harvard University Press.
Pelikan, P., & Wagner, G. (Eds.). (2003). The evolutionary analysis of economic policy. Chel-
tenham, UK; Northampton, MA: Edward Elgar.
Pies, I. (2000). Ordnungspolitik in der Demokratie: ein ökonomischer Ansatz diskursiver Politik-
beratung. Tübingen: Mohr Siebeck.
Pohl, R., & Hell, W. (1996). No reduction in hindsight bias after complete information and repeated
testing, Organizational Behavior on Human Decision Processes, 67(1), 49–58.
Polya, G. (1954). Mathematics and plausible reasoning. Vol. 1: Induction and analogy in mathe-
matics [Gigerenzer/Todd (1999)]. Princeton, NJ: Princeton University Press.
Literature 203
Pombeni, F. (February, 2005). Debiasing-Problematik, Methoden und eine experimentelle Fall-

studie. Dissertation submitted for a diploma at the Technische Universität Dresden under the
supervision of Prof. Dr. M. Lehmann-Waffenschmidt.
Popper, K. (1957). The poverty of historicism. New York: Harper and Row.
Popper, K. (1967). Unended quest: An intellectual autobiography. London: Fontana.
Preston, M. G., & Baratta, P. (1948). An experimental study of the auction value of an uncertain
outcome. American Journal of Psychology, 61:183–93.
Priest, G. (1987). Unstable solutions to the liar paradox. In S. J. Bartlett,, P. Suber (Eds.), Self-
reference. Reflections on reflexivity. Dordrecht: Martinus Nijhoff Publishers.
Pruitt, D. G. (1981). Negotiation behavior. New York: Academic Press.
Quine, W. V. (1962). Paradox. In S. J. Bartlett (Ed.), Reflexivity. A source-book in self-reference
[Reprinted form Scientific American, 20(4), 84–96.]. North-Holland: Elsevier Science Publish-
ers B. V.
Rabin, M. (1998). Psychology and economics. Journal of Economic Literature, 36(1), 11–46.
Raiffa, H. (1982). The art and science of negotiation. Cambridge: Harvard University Press.
Rakow, T., Harvey, N., & Finer, S. (2003). Improving calibration without training: The Role of
task information. Applied Cognitive Psychology, 17, 419–441.
Reichenbach, H. (1947). In elements of symbolic logic. New York: The Free Press.
Rosenthal, R. (1998). Covert communication in classrooms, clinics, and courtrooms. Eye on Psi
Chi, 3(1), 18–22.
Rosenthal, R., & Jacobson, L. (1968/1992). Pygmalion in the classroom: Teacher expectation and
pupils’ intellectual development. New York: Irvington Publishers.
Roshwald, M. (1955). Value-judgements in the social sciences. British Journal for the Philosophy
of Science, 6(23), 186–208.
Roth, A. E. (1995). Bargaining experiments. In J. H. Kagel, & A. Roth, (Eds.). The handbook of
experimental economics. Princeton: Princeton University Press.
Roth, A., & Erev, I. (1995). Learning in extensive-form games: experimental data and simple dy-
namic models in the intermediate term. Games and Economic Behavior, 8(Special Issue: Nobel
Symposium), 164–212.
Roth, G. (1985a). Die Konstruktivität des Gehirns: Der Kenntnisstand der Hirnforschung. In
H. R. Fischer (Ed.), Die Wirklichkeit des KOnstruktivisms – Zur Auseinandersetzung um ein
neues Paradigma (pp. 47–71). Heidelberg: Carl-Auer-Systeme-Verlag.
Roth, G. (1985b). Die Selbstreferentialität des Gehirns und die Prinzipien der Gestalt-
wahrnehmung, Gestalt Theory, 7, 228–244.
Rothstein, B. (2005). Social traps and the problem of trust. Cambridge: Cambridge Univer-
sity Press.
Roy, M. C., & Lerch, F. J. (1996). Overcoming ineffective mental representations in base-rate
problems. Information System Research, 7(2), 16–25.
Rubinstein, A., Tversky, A., & Heller, D. (1997). Naive strategies in competitive games. In Albers,
W., Güth, W., Hammerstein, P., Moldovanu, B., Van Damme, E., Selten, R. (Eds.), Under-
standing strategic interaction: Essays in honor of Reinhard Selten (pp. 394–402). Heidelberg:
Springer.
Rusch, G. (Ed.). (1999). Wissen und Wirklichkeit – Beiträge zum Konstruktivismus – Eine Hom-
mage an Hernst von Glasersfeld, Heidelberg: Carl-Auer-Systeme Verlag.
Russell, B. (1903). Principles of mathematics. Cambridge: Cambridge University Press.
Russell, B. (1940). An inquiry into meaning and truth. London: Allen & Unwin.
Russo, J. E., & Schoemacher, P. J. H. (1992, Winter). Managing overconfidence, Sloan Manage-
ment Review, 7–17.
Salewski, M. (1999). Was wäre wenn. Alternativ und Parallelgeschichte: Brücken zwischen Phan-
tasie und Wirklichkeit. Steiner: Stuttgart.
Sanna, L. J., & Schwarz, N. (2004). Integrating temporal biases: The interplay of focal thoughts
and accessibility experiences. Psychological Science, 15, 474–481.
Sanna, L. J., Stocker, S. L., & Schwarz, N. (2002). When debiasing backfires: Accessible con-
tent and accessibility experiences in debiasing hindsight. Journal of Experimental Psychology,
Learning, Memory and Cognition, 28(3), 497–502.
204 Literature
Sapir, E. (1929). The status of linguistics as a science. In E. Sapir, & D. G. Mandelbaum (Eds.),
Selected writings of edward sapir in language, culture, and personality. Berkeley: University
of California Press.
Sapir, E., & Mandelbaum, D. G. (Eds.). (1986). Selected writings of edward sapir in language,
culture, and personality. Berkeley: University of California Press.
Sbriglia, P. (2008). Revealing the depth of reasoning in p-beauty contest games, Experimental
Economics, 11(2), 107–202.
Scharfstein, D. S., & Stein, J. C. (1990). Herd behaviour and investment. American Economic
Review, 80, 465–479.
Scheutz, M. (1995). Ist das der Titel eines Buchs? Selbstreferenz neu analysiert. Dissertationen der
Universität Wien, Band 13, WUV-Universitätsverlag.
Schmidt, S. J. (1987). Der Diskurs des Radikalen Konstruktivismus. Frankfurt am Main: Suhrkamp
Verlag.
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information processing:
1. Detection, search, and attention. Psychological Review, 84, 1–66.
Schrödinger, E. (1935, November). Die gegenwärtige Situation in der Quantenmechanik. Natur-
wissenschaften.
Schumpeter, J. A. (1954). History of Economic Analysis [Revised arch 1996], USA: Oxford
University Press.
Schwegler, H. (1999). Wissenschaft als Sprachspiel. In G. Rusch (Ed.), Wissen und Wirklichkeit –
Beiträge zum Konstruktivismus – Eine Hommage an Hernst von Glasersfeld (pp. 17–34),
Heidelberg: Carl-Auer-Systeme Verlag.
Scott, R. W. (1995). Institutions and organizations. Thousand Oaks, CA: Sage.
Searle, J. (1995). The construction of social reality. NewYork: Free Press.
Sebenius, J. K. (1992). Negotiation analysis: A characterization and review. Management Science,
38(1), 18–38.
Selten, R. (1991). Anticipatory Learning in Two-Person Games, in Selten, R. (Ed.), Game equilib-
rium models I. Berlin: Springer Verlag, 1991, pp. 98–154.
Selten, R. (1999). What is Bounded Rationality? Paper prepared for the Dahlem Conference,
Mimeo.
Shiller, R. (2001, May). Bubbles human judgment and expert opinion. Cowles Foundation Discus-
sion Paper No. 1303.
Simon, H. (1955). A behavioral model of rational choice. Quarterly Journal of Economics, 69,
99–118.
Simon, H. (1956a). Rational choice and the structure of the environment. In H. Simon, M. Egidi,
R. Marris, & R. Viale (Eds.), Economics, bounded rationality and the cognitive revolution,
Northampton, MA: Edward Elgar Publishing Company.
Simon, H. (1956b). Rational choice and the structure of the environments. Psychological Review,
63, 129–138.
Simon, H. (1957). Models of man. New York: Wiley.
Simon, H. (1977). Models of discovery and other topics in the methods of science. Boston:
D. D. Riedle Publishing Company.
Simon, H. (1981). The sciences of artificial. Cambridge, MA: MIT Press.
Simon, H. (1982). Models of bounded rationality. Cambridge, MA: MIT Press.
Simon, H. (1988). Scientific discovery as problem solving. In H. E. Simon, M. Egidi, R. Marris,
& R. Viale (Eds.), Economics, bounded rationality and the cognitive revolution. Northampton,
MA: Edward Elgar Publishing Company.
Simon, H. (1990). Invariants of human behaviour. Annual Review of Psychology, 41, 1–19.
Simon, H. (1992). Introductory comments. In H. Simon, M. Egidi, R. Marris, & R. Viale (Eds.),
Economics, bounded rationality and the cognitive revolution (pp. 3–7). Northampton, MA:
Edward Elgar Publishing Company.
Simon, H., Hayes, J. R. (1976). The understanding process: problem isomorphs. Cognotive Psy-
chology, 8, 165–190.
Literature 205
Sipser, M. (2006). Introduction to the theory of computation. Boston, MA: Thompson Course
Technology Division of Thompson Learning, Inc.
Slembeck, T. (2003). Ideologies, beliefs and economic advice – A cognitive-evolutionary view
on economic policy-making. In P. Pelikan, & G. Wagner (Eds.), The evolutionary analysis of
economic policy, Cheltenham, UK; Northampton, MA, USA: Edward Elgar.
Smith, K. K., & Crandall, S. D. (1984). Exploring collective emotions. American Behavioral Sci-
entist, 27, 813–828.
Smith, K. K., Simmon, V. M., & Thames, T. B. (1989). ‘Fix the Women’: An intervention into
an organizational conflict based on parallel process thinking. Journal of Applied Behavioral
Science, 25, 11–29.
Smith, V. (1976). Experimental economics: induced value theory. American Economic Review, 66,
274–279.
Smith, V. (1982). Microeconomic systems as an experimental science. The American Economic
Review, 923–955.
Smullyan, R. (1991). Godel’s incompleteness theorems. New York: Oxford University Press.
Smullyan, R. M. (1984). Chameleonic languages. Synthese, 60(2), 201–224.
Solé, R. V., & Bascompte, J. (2006). Self organization in complex ecosystems. Princeton, NJ:
Princeton University Press.
Soros, G. (1994). The Alchemy of finance – Reading the mind of the market. New Work: Wiley.
Stahl, D. O. (1993). The Evolution of Smart Players. Games and Economic Behavior, October
1993, 5(4), 604–17.
Stahl, D. O., & Wilson, P. W. (1994). Experimental Evidence on Players’ Models of Other Players.
Journal of Economic Behavior and Organization, December 1994, 25(3), pp. 309–27.
Stahl, D. O. (1995). Boundedly rational rule learning in a guesssing game. Games and Economic
Behavior, 16, 303–330.
Stanovich, K. E., & West, R. F. (2002). Individual differences in reasoning: Implications for the
rationality debate. In T. Gilovich, D.Griffin, & D. Kahneman (Eds.), Heuristics and biases: The
psychology of intuitive judgment (pp. 421–440). Cambridge, UK: Cambridge University Press.
Stapel, D. A., Reicher, S. D., & Spears, R. (1995). Contextual determinants of strategic choice:
Some moderators of the availability bias. European Journal of Social Psychology, 25, 141–158.
Steier, F. (1991a). Introduction: Research as reflexivity, self-reflexivity as social process. In F.
Steier (Ed.), Research as reflexivity (pp. 1–11). California: Sage Publications Ltd.
Steier, F. (1991b). Reflexivity and methodology: An ecological constructionism. In F. Steier, (Ed.),
Research as reflexivity(pp. 163–185). California: Sage Publications Ltd.
Stiehler, A. (2003). Do individuals recognize cascade behavior of others? – An experimental study.
Discussion Paper, 02, Max Planck Institute for Research into Economic Systems, Jena.
Stigler, G. J. (1961). The economics of information. Journal of Political Economy, 69, 213–225.
Stratmann, T. (1997). Logrolling, In D. C. Mueller, (Ed.), Perspectives on public choice: A hand-
book. Cambridge, UK: Cambridge University Press.
Suber, P. (1987b). Varieties of self-reference. In S. J. Bartlett, & P. Suber (Eds.), Self-reference.
Reflections on reflexivity. Dordrecht: Martinus Nijhoff Publishers.
Suber, P. (1989). The reflexivity of change: The case of language norms. Journal of Speculative
Philosophy, 3(2), 100–129.
Suber, P. (1990). The paradox of self-amendment: A study of logic, law, omnipotence, and change.
New York: Peter Lang Publishing.
Tamborini, R. (1997). Constructivism. A pattern of human knowledge for economics. In R. Viale,
(Ed.), Cognitive economics. LaSCoMES Series, I/1997.
Thomas, W. I., & Thomas, D. (1928). The child in America. New York: Knopf.
Thurston, L. L. (1931). The indifference function. Journals of Social Psychology, 2:39-67.
Tietzel, M. (1989). Prognoselogic, oder: Warum Prognostiker irren dürfen – On the logic of eco-
nomic forecasting. Jahrbuch für Nationalökonomie und Statistik, 206(6), 246–262.
Todd, P. M., & Miller, G. F. (1999). From pride and prejudice to persuasion: Satisficing in mate
search. In G. Gigerenzer, P. Todd, & the ABC Research Group (Eds.), Simple heuristics that
make us smart (pp. 287–308). Oxford: Oxford University Press.
206 Literature
Tversky, A., & Kahneman, D. (1980). Causal schemas in judgement under uncertainty. In
M. Fischbein, (Ed.), Progress in social psychology. Hillsdale, NJ: Erlbaum.
Tversky, A., & Kahnemann, D. (1982). Judgement under uncertainty: Heuristics and biases.
In D. Kahnemann, P. Slovic, A. Tversky (Eds.), Judgement under uncertainty (pp. 3–21).
Cambridge: Cambridge University Press.
Van Fraassen, B. C. (1970, October). Inference and self-reference. Synthese, 21(3–4).
Venn (1966). The logic of change (4th ed.). New York: New York Chelsea Publishing Co.
von Foerster, H. (1992). Entdecken oder Erfinden – Wie lässt sich Verstehen verstehen? In
G. Heinz, & H. Meier, (Eds.), Einführung in den Konstruktivismus (pp. 41–88), 7. Aufl. (1.
Aufl. 1992).
Von Foerster, H., & Von Glasersfeld, E. (1999). Wie wir uns erfinden. Eine Autobiographie des
radikalen Konstruktivismus. Heidelberg: Carl Auer.
Von Glasersfeld, E. (1996). Radikaler Konstruktivismus – Ideen, Ergebnisse, Probleme [Radical
constructivism. A way of knowing and learning, London: The Falmer Press, 1995]. Frankfurt
am Main: Suhrkamp.
Von Humboldt, W. (1945). Über das vergleichende Sprachstudium in Beziehung auf die ver-
schiedenen Epochen der Sprachentwicklung. Leipzig: Felix Meiner.
Von Neumann, J., & Morgenstern, O. (1944). Theory of games and economic behavior. Düsseldorf:
Verlag Wirtschaft und Finanzen.
Watzlawick, P. (Ed.). (1981). Die erfundene Wirklichkeit – Was wissen wir, was wir zu wissen
glauben? – Beiträge zum Konstuktivismus. München: R.Piper & Co. Verlag.
Weber, R. A. (2003). Learning with no feedback in a competitive guessing game. Games and
Economic Behavior, 44, 134–144.
Wehner, B. (1995). Die Logik der Politik und das Elend der Ökonomie. Darmstadt.
Weick, K. E. (1995). Sensemaking in organisations. Thousand Oaks, CA: Sage.
Weiss, P. (1992). The theory of type. In S. J. Bartlett (Ed.), Reflexivity. A source-book in self-
reference. North-Holland: Elsevier Science Publishers B. V.
Welch, I. (2000). Herding among security analysts. Journal of Financial Economics, 58(3),
369–396.
Wellman, H. M. (1990). The child’s theory of mind, Cambridge, MA: MIT Press.
Whewell (1987). Self-reference and meaning in the natural language. In S. J. Bartlett, & P. Suber
(Eds.), Self-reference. Reflections on reflexivity. Dordrecht: Martinus Nijhoff Publishers.
Whitehead, A. N., Russell, B. (1910, 1912, 1913). Principia Mathematica [2nd ed., 1925 (Vol. 1),
1927 (Vols 2, 3). Abridged as: Principia Mathematica to ∗ 56, abridged, Cambridge University
Press, 1962]. Cambridge: Cambridge University Press.
Whorf, B., & Carroll, J. (Eds.) (1964). Language, thought, and reality: Selected writings of ben-
jamin lee whorf. Cambridge, MA: MIT Press.
Willinger, M., & Ziegelmeyer, A. (1998). Are more informed agents able to shatter information
cascades in the lab? In P. Cohendet, P. Llerena, H. Stahn, & G. Umbhauer (Eds.), The economics
of networks: Interaction and behaviours (pp. 291–305), Heidelberg: Springer.
Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function
of wrong beliefs in young children’s understanding of deception. Cognition, 13, 103–128.
Winrich, J. S. (1984, December). Self-reference and the incomplete structure of neoclassical eco-
nomics. Journal of Economic Issues, 18(4).
Wittgenstein, L. (1953). Philosophische Untersuchungen. Bibliothek Suhrkamp.
Wittgenstein, L. (1999). Tractatus logico-philosophicus. Tagebücher 1914–1916. Philosophische
Untersuchungen (Vol. 1). Frankfurt am Main: Suhrkamp.
Woolgar (1992). Some remarks about positionism: A peply to collins and yearly. In A. Pickering,
(Ed.), Science as practice and culture (pp. 327–342). Chicago: University of Chicago.
View publication stats

Reflexivity in Economics 2009

Uploaded by

Copyright:

Available Formats

Reflexivity in Economics 2009

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reflexivity in Economics 2009

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Reﬂexivity in economics. An experimental examination on the self-

Book · January 2009

The user has requested enhancement of the downloaded file.

ISBN: 978-3-7908-2091-1 e-ISBN: 978-3-7908-2092-8

Contributions to Economics ISSN: 1431-1933

Library of Congress Control Number: 2008935108

© 2009 Physica-Verlag Heidelberg

Cover design: WMXDesign GmbH, Heidelberg

Printed on acid-free paper

I would like to thank my supervisor, Prof. Marco Lehmann-Waffenschmidt, for his

1 Reflexivity and Self-Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Reflexivity of Social Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Reflexivity and Predictability of the Social Sciences . . . . . . . . . . . . . . . . 53

4 On the Rationality of the Economic Actors . . . . . . . . . . . . . . . . . . . . . . . . 77

5 Heuristics, Biases and Methods for Debiasing . . . . . . . . . . . . . . . . . . . . . 101

6 Self-Referentiality of Economic Theories and Theory Absorption . . . . 119

7 On the Absorbability of Economic Theories – An Experimental

The investigation of the notion of “self-reference” will, in particular, indicate that

1 Cf. Bartlett (1987, p. 5).

Reflexivity is “the quality or state of being reflexive,”3 whereas “reflexive” stems

3 From the Merriam-Webster’s Unabridged Dictionary (2000).

as something which aims at depicting or characterising itself. Therefore, they es-

1.2 Reference Relations

A reference relation R connects a class of referring objects with a class of objects

14 Cf. Davis and Klaes (2003, p. 333).

Fig. 1.2 Reference relations

Fig. 1.3 Reference relations for S1 and S2 (Bolander, 2002)

17 Cf. Bolander (2002, p. 10 ff).

18 Cf. Bolander (2002, p. 18).

is because, as stated by the theory of types, “it is impossible or meaningless to state

Finally, self-reference is likely to appear in situations that show an ungrounded na-

1.3 Varieties of Self-Reference

It is possible to enumerate many types of self-references and several classifications

22 Cf. Weiss (1992, p. 37).

25 The main reference for this paragraph is Bartlett (1987).

31 Cf. Peirce (1931–58).

a possibility to escape the self-referentiality of defining and of dictionary attribu-

1.4 Taxonomies of Self-References

Basically, it can be distinguished between self-references which establish circular

1.5 Self-Reference in Natural Language

In examining self-reference in grammar, Van Fraassen47 distinguishes between two

43 Cf. Bartlett (1987, p. 10).

difference among them lies in having a variable (i.e. “context dependent”), or an

48 Cf. Van Fraassen (1970, p. 696).

Whenever a sentence type is self-referring, i.e. whenever each of its token is

55 Cf. Martin (1992, p. 133).

necessary “to go outside the statement in order to know whether it is self-referring

1.6 Self-Reference in Formal Language

A formal language68 (L) is defined by mathematical or machine processable formu-

23th January 2007.

formalized language system (interpreted language).69 This could be of a certain in-

1.7 Logical Consistency of Self-Reference

1.7.1 Self-Reference and Paradoxes

The concept of a “paradox” can be applied to statements that appear plausi-

formulated it in 1908. Cf. Quine (1962).

1.7.2 Harmless and Harmful Self-Reference

81 Cf. Merriam-Webster’s Unabridged Dictionary (2000) at the voice “exception.”

On the contrary, reflexivity which acts in a self-refuting way, based on a negative